]>
Commit | Line | Data |
---|---|---|
6cbd5570 CM |
1 | /* |
2 | * Copyright (C) 2007 Oracle. All rights reserved. | |
3 | * | |
4 | * This program is free software; you can redistribute it and/or | |
5 | * modify it under the terms of the GNU General Public | |
6 | * License v2 as published by the Free Software Foundation. | |
7 | * | |
8 | * This program is distributed in the hope that it will be useful, | |
9 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
10 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
11 | * General Public License for more details. | |
12 | * | |
13 | * You should have received a copy of the GNU General Public | |
14 | * License along with this program; if not, write to the | |
15 | * Free Software Foundation, Inc., 59 Temple Place - Suite 330, | |
16 | * Boston, MA 021110-1307, USA. | |
17 | */ | |
18 | ||
e20d96d6 | 19 | #include <linux/fs.h> |
d98237b3 | 20 | #include <linux/blkdev.h> |
87cbda5c | 21 | #include <linux/scatterlist.h> |
22b0ebda | 22 | #include <linux/swap.h> |
0f7d52f4 | 23 | #include <linux/radix-tree.h> |
35b7e476 | 24 | #include <linux/writeback.h> |
d397712b | 25 | #include <linux/buffer_head.h> |
ce9adaa5 | 26 | #include <linux/workqueue.h> |
a74a4b97 | 27 | #include <linux/kthread.h> |
5a0e3ad6 | 28 | #include <linux/slab.h> |
784b4e29 | 29 | #include <linux/migrate.h> |
7a36ddec | 30 | #include <linux/ratelimit.h> |
6463fe58 | 31 | #include <linux/uuid.h> |
803b2f54 | 32 | #include <linux/semaphore.h> |
7e75bf3f | 33 | #include <asm/unaligned.h> |
eb60ceac CM |
34 | #include "ctree.h" |
35 | #include "disk-io.h" | |
0b947aff | 36 | #include "hash.h" |
e089f05c | 37 | #include "transaction.h" |
0f7d52f4 | 38 | #include "btrfs_inode.h" |
0b86a832 | 39 | #include "volumes.h" |
db94535d | 40 | #include "print-tree.h" |
925baedd | 41 | #include "locking.h" |
e02119d5 | 42 | #include "tree-log.h" |
fa9c0d79 | 43 | #include "free-space-cache.h" |
70f6d82e | 44 | #include "free-space-tree.h" |
581bb050 | 45 | #include "inode-map.h" |
21adbd5c | 46 | #include "check-integrity.h" |
606686ee | 47 | #include "rcu-string.h" |
8dabb742 | 48 | #include "dev-replace.h" |
53b381b3 | 49 | #include "raid56.h" |
5ac1d209 | 50 | #include "sysfs.h" |
fcebe456 | 51 | #include "qgroup.h" |
ebb8765b | 52 | #include "compression.h" |
557ea5dd | 53 | #include "tree-checker.h" |
fd708b81 | 54 | #include "ref-verify.h" |
eb60ceac | 55 | |
de0022b9 JB |
56 | #ifdef CONFIG_X86 |
57 | #include <asm/cpufeature.h> | |
58 | #endif | |
59 | ||
319e4d06 QW |
60 | #define BTRFS_SUPER_FLAG_SUPP (BTRFS_HEADER_FLAG_WRITTEN |\ |
61 | BTRFS_HEADER_FLAG_RELOC |\ | |
62 | BTRFS_SUPER_FLAG_ERROR |\ | |
63 | BTRFS_SUPER_FLAG_SEEDING |\ | |
64 | BTRFS_SUPER_FLAG_METADUMP) | |
65 | ||
e8c9f186 | 66 | static const struct extent_io_ops btree_extent_io_ops; |
8b712842 | 67 | static void end_workqueue_fn(struct btrfs_work *work); |
4df27c4d | 68 | static void free_fs_root(struct btrfs_root *root); |
3d3a126a | 69 | static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info); |
143bede5 | 70 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root); |
acce952b | 71 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
2ff7e61e | 72 | struct btrfs_fs_info *fs_info); |
143bede5 | 73 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); |
2ff7e61e | 74 | static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, |
acce952b | 75 | struct extent_io_tree *dirty_pages, |
76 | int mark); | |
2ff7e61e | 77 | static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, |
acce952b | 78 | struct extent_io_tree *pinned_extents); |
2ff7e61e JM |
79 | static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info); |
80 | static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info); | |
ce9adaa5 | 81 | |
d352ac68 | 82 | /* |
97eb6b69 DS |
83 | * btrfs_end_io_wq structs are used to do processing in task context when an IO |
84 | * is complete. This is used during reads to verify checksums, and it is used | |
d352ac68 CM |
85 | * by writes to insert metadata for new file extents after IO is complete. |
86 | */ | |
97eb6b69 | 87 | struct btrfs_end_io_wq { |
ce9adaa5 CM |
88 | struct bio *bio; |
89 | bio_end_io_t *end_io; | |
90 | void *private; | |
91 | struct btrfs_fs_info *info; | |
4e4cbee9 | 92 | blk_status_t status; |
bfebd8b5 | 93 | enum btrfs_wq_endio_type metadata; |
8b712842 | 94 | struct btrfs_work work; |
ce9adaa5 | 95 | }; |
0da5468f | 96 | |
97eb6b69 DS |
97 | static struct kmem_cache *btrfs_end_io_wq_cache; |
98 | ||
99 | int __init btrfs_end_io_wq_init(void) | |
100 | { | |
101 | btrfs_end_io_wq_cache = kmem_cache_create("btrfs_end_io_wq", | |
102 | sizeof(struct btrfs_end_io_wq), | |
103 | 0, | |
fba4b697 | 104 | SLAB_MEM_SPREAD, |
97eb6b69 DS |
105 | NULL); |
106 | if (!btrfs_end_io_wq_cache) | |
107 | return -ENOMEM; | |
108 | return 0; | |
109 | } | |
110 | ||
111 | void btrfs_end_io_wq_exit(void) | |
112 | { | |
5598e900 | 113 | kmem_cache_destroy(btrfs_end_io_wq_cache); |
97eb6b69 DS |
114 | } |
115 | ||
d352ac68 CM |
116 | /* |
117 | * async submit bios are used to offload expensive checksumming | |
118 | * onto the worker threads. They checksum file and metadata bios | |
119 | * just before they are sent down the IO stack. | |
120 | */ | |
44b8bd7e | 121 | struct async_submit_bio { |
c6100a4b JB |
122 | void *private_data; |
123 | struct btrfs_fs_info *fs_info; | |
44b8bd7e | 124 | struct bio *bio; |
4a69a410 CM |
125 | extent_submit_bio_hook_t *submit_bio_start; |
126 | extent_submit_bio_hook_t *submit_bio_done; | |
44b8bd7e | 127 | int mirror_num; |
c8b97818 | 128 | unsigned long bio_flags; |
eaf25d93 CM |
129 | /* |
130 | * bio_offset is optional, can be used if the pages in the bio | |
131 | * can't tell us where in the file the bio should go | |
132 | */ | |
133 | u64 bio_offset; | |
8b712842 | 134 | struct btrfs_work work; |
4e4cbee9 | 135 | blk_status_t status; |
44b8bd7e CM |
136 | }; |
137 | ||
85d4e461 CM |
138 | /* |
139 | * Lockdep class keys for extent_buffer->lock's in this root. For a given | |
140 | * eb, the lockdep key is determined by the btrfs_root it belongs to and | |
141 | * the level the eb occupies in the tree. | |
142 | * | |
143 | * Different roots are used for different purposes and may nest inside each | |
144 | * other and they require separate keysets. As lockdep keys should be | |
145 | * static, assign keysets according to the purpose of the root as indicated | |
146 | * by btrfs_root->objectid. This ensures that all special purpose roots | |
147 | * have separate keysets. | |
4008c04a | 148 | * |
85d4e461 CM |
149 | * Lock-nesting across peer nodes is always done with the immediate parent |
150 | * node locked thus preventing deadlock. As lockdep doesn't know this, use | |
151 | * subclass to avoid triggering lockdep warning in such cases. | |
4008c04a | 152 | * |
85d4e461 CM |
153 | * The key is set by the readpage_end_io_hook after the buffer has passed |
154 | * csum validation but before the pages are unlocked. It is also set by | |
155 | * btrfs_init_new_buffer on freshly allocated blocks. | |
4008c04a | 156 | * |
85d4e461 CM |
157 | * We also add a check to make sure the highest level of the tree is the |
158 | * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code | |
159 | * needs update as well. | |
4008c04a CM |
160 | */ |
161 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
162 | # if BTRFS_MAX_LEVEL != 8 | |
163 | # error | |
164 | # endif | |
85d4e461 CM |
165 | |
166 | static struct btrfs_lockdep_keyset { | |
167 | u64 id; /* root objectid */ | |
168 | const char *name_stem; /* lock name stem */ | |
169 | char names[BTRFS_MAX_LEVEL + 1][20]; | |
170 | struct lock_class_key keys[BTRFS_MAX_LEVEL + 1]; | |
171 | } btrfs_lockdep_keysets[] = { | |
172 | { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" }, | |
173 | { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" }, | |
174 | { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" }, | |
175 | { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" }, | |
176 | { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" }, | |
177 | { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" }, | |
60b62978 | 178 | { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" }, |
85d4e461 CM |
179 | { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" }, |
180 | { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" }, | |
181 | { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" }, | |
13fd8da9 | 182 | { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" }, |
6b20e0ad | 183 | { .id = BTRFS_FREE_SPACE_TREE_OBJECTID, .name_stem = "free-space" }, |
85d4e461 | 184 | { .id = 0, .name_stem = "tree" }, |
4008c04a | 185 | }; |
85d4e461 CM |
186 | |
187 | void __init btrfs_init_lockdep(void) | |
188 | { | |
189 | int i, j; | |
190 | ||
191 | /* initialize lockdep class names */ | |
192 | for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) { | |
193 | struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i]; | |
194 | ||
195 | for (j = 0; j < ARRAY_SIZE(ks->names); j++) | |
196 | snprintf(ks->names[j], sizeof(ks->names[j]), | |
197 | "btrfs-%s-%02d", ks->name_stem, j); | |
198 | } | |
199 | } | |
200 | ||
201 | void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, | |
202 | int level) | |
203 | { | |
204 | struct btrfs_lockdep_keyset *ks; | |
205 | ||
206 | BUG_ON(level >= ARRAY_SIZE(ks->keys)); | |
207 | ||
208 | /* find the matching keyset, id 0 is the default entry */ | |
209 | for (ks = btrfs_lockdep_keysets; ks->id; ks++) | |
210 | if (ks->id == objectid) | |
211 | break; | |
212 | ||
213 | lockdep_set_class_and_name(&eb->lock, | |
214 | &ks->keys[level], ks->names[level]); | |
215 | } | |
216 | ||
4008c04a CM |
217 | #endif |
218 | ||
d352ac68 CM |
219 | /* |
220 | * extents on the btree inode are pretty simple, there's one extent | |
221 | * that covers the entire device | |
222 | */ | |
fc4f21b1 | 223 | static struct extent_map *btree_get_extent(struct btrfs_inode *inode, |
306e16ce | 224 | struct page *page, size_t pg_offset, u64 start, u64 len, |
b2950863 | 225 | int create) |
7eccb903 | 226 | { |
fc4f21b1 NB |
227 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->vfs_inode.i_sb); |
228 | struct extent_map_tree *em_tree = &inode->extent_tree; | |
5f39d397 CM |
229 | struct extent_map *em; |
230 | int ret; | |
231 | ||
890871be | 232 | read_lock(&em_tree->lock); |
d1310b2e | 233 | em = lookup_extent_mapping(em_tree, start, len); |
a061fc8d | 234 | if (em) { |
0b246afa | 235 | em->bdev = fs_info->fs_devices->latest_bdev; |
890871be | 236 | read_unlock(&em_tree->lock); |
5f39d397 | 237 | goto out; |
a061fc8d | 238 | } |
890871be | 239 | read_unlock(&em_tree->lock); |
7b13b7b1 | 240 | |
172ddd60 | 241 | em = alloc_extent_map(); |
5f39d397 CM |
242 | if (!em) { |
243 | em = ERR_PTR(-ENOMEM); | |
244 | goto out; | |
245 | } | |
246 | em->start = 0; | |
0afbaf8c | 247 | em->len = (u64)-1; |
c8b97818 | 248 | em->block_len = (u64)-1; |
5f39d397 | 249 | em->block_start = 0; |
0b246afa | 250 | em->bdev = fs_info->fs_devices->latest_bdev; |
d1310b2e | 251 | |
890871be | 252 | write_lock(&em_tree->lock); |
09a2a8f9 | 253 | ret = add_extent_mapping(em_tree, em, 0); |
5f39d397 CM |
254 | if (ret == -EEXIST) { |
255 | free_extent_map(em); | |
7b13b7b1 | 256 | em = lookup_extent_mapping(em_tree, start, len); |
b4f359ab | 257 | if (!em) |
0433f20d | 258 | em = ERR_PTR(-EIO); |
5f39d397 | 259 | } else if (ret) { |
7b13b7b1 | 260 | free_extent_map(em); |
0433f20d | 261 | em = ERR_PTR(ret); |
5f39d397 | 262 | } |
890871be | 263 | write_unlock(&em_tree->lock); |
7b13b7b1 | 264 | |
5f39d397 CM |
265 | out: |
266 | return em; | |
7eccb903 CM |
267 | } |
268 | ||
9ed57367 | 269 | u32 btrfs_csum_data(const char *data, u32 seed, size_t len) |
19c00ddc | 270 | { |
0b947aff | 271 | return btrfs_crc32c(seed, data, len); |
19c00ddc CM |
272 | } |
273 | ||
0b5e3daf | 274 | void btrfs_csum_final(u32 crc, u8 *result) |
19c00ddc | 275 | { |
7e75bf3f | 276 | put_unaligned_le32(~crc, result); |
19c00ddc CM |
277 | } |
278 | ||
d352ac68 CM |
279 | /* |
280 | * compute the csum for a btree block, and either verify it or write it | |
281 | * into the csum field of the block. | |
282 | */ | |
01d58472 DD |
283 | static int csum_tree_block(struct btrfs_fs_info *fs_info, |
284 | struct extent_buffer *buf, | |
19c00ddc CM |
285 | int verify) |
286 | { | |
01d58472 | 287 | u16 csum_size = btrfs_super_csum_size(fs_info->super_copy); |
607d432d | 288 | char *result = NULL; |
19c00ddc CM |
289 | unsigned long len; |
290 | unsigned long cur_len; | |
291 | unsigned long offset = BTRFS_CSUM_SIZE; | |
19c00ddc CM |
292 | char *kaddr; |
293 | unsigned long map_start; | |
294 | unsigned long map_len; | |
295 | int err; | |
296 | u32 crc = ~(u32)0; | |
607d432d | 297 | unsigned long inline_result; |
19c00ddc CM |
298 | |
299 | len = buf->len - offset; | |
d397712b | 300 | while (len > 0) { |
19c00ddc | 301 | err = map_private_extent_buffer(buf, offset, 32, |
a6591715 | 302 | &kaddr, &map_start, &map_len); |
d397712b | 303 | if (err) |
8bd98f0e | 304 | return err; |
19c00ddc | 305 | cur_len = min(len, map_len - (offset - map_start)); |
b0496686 | 306 | crc = btrfs_csum_data(kaddr + offset - map_start, |
19c00ddc CM |
307 | crc, cur_len); |
308 | len -= cur_len; | |
309 | offset += cur_len; | |
19c00ddc | 310 | } |
607d432d | 311 | if (csum_size > sizeof(inline_result)) { |
31e818fe | 312 | result = kzalloc(csum_size, GFP_NOFS); |
607d432d | 313 | if (!result) |
8bd98f0e | 314 | return -ENOMEM; |
607d432d JB |
315 | } else { |
316 | result = (char *)&inline_result; | |
317 | } | |
318 | ||
19c00ddc CM |
319 | btrfs_csum_final(crc, result); |
320 | ||
321 | if (verify) { | |
607d432d | 322 | if (memcmp_extent_buffer(buf, result, 0, csum_size)) { |
e4204ded CM |
323 | u32 val; |
324 | u32 found = 0; | |
607d432d | 325 | memcpy(&found, result, csum_size); |
e4204ded | 326 | |
607d432d | 327 | read_extent_buffer(buf, &val, 0, csum_size); |
94647322 | 328 | btrfs_warn_rl(fs_info, |
5d163e0e | 329 | "%s checksum verify failed on %llu wanted %X found %X level %d", |
01d58472 | 330 | fs_info->sb->s_id, buf->start, |
efe120a0 | 331 | val, found, btrfs_header_level(buf)); |
607d432d JB |
332 | if (result != (char *)&inline_result) |
333 | kfree(result); | |
8bd98f0e | 334 | return -EUCLEAN; |
19c00ddc CM |
335 | } |
336 | } else { | |
607d432d | 337 | write_extent_buffer(buf, result, 0, csum_size); |
19c00ddc | 338 | } |
607d432d JB |
339 | if (result != (char *)&inline_result) |
340 | kfree(result); | |
19c00ddc CM |
341 | return 0; |
342 | } | |
343 | ||
d352ac68 CM |
344 | /* |
345 | * we can't consider a given block up to date unless the transid of the | |
346 | * block matches the transid in the parent node's pointer. This is how we | |
347 | * detect blocks that either didn't get written at all or got written | |
348 | * in the wrong place. | |
349 | */ | |
1259ab75 | 350 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
b9fab919 CM |
351 | struct extent_buffer *eb, u64 parent_transid, |
352 | int atomic) | |
1259ab75 | 353 | { |
2ac55d41 | 354 | struct extent_state *cached_state = NULL; |
1259ab75 | 355 | int ret; |
2755a0de | 356 | bool need_lock = (current->journal_info == BTRFS_SEND_TRANS_STUB); |
1259ab75 CM |
357 | |
358 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
359 | return 0; | |
360 | ||
b9fab919 CM |
361 | if (atomic) |
362 | return -EAGAIN; | |
363 | ||
a26e8c9f JB |
364 | if (need_lock) { |
365 | btrfs_tree_read_lock(eb); | |
366 | btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK); | |
367 | } | |
368 | ||
2ac55d41 | 369 | lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, |
ff13db41 | 370 | &cached_state); |
0b32f4bb | 371 | if (extent_buffer_uptodate(eb) && |
1259ab75 CM |
372 | btrfs_header_generation(eb) == parent_transid) { |
373 | ret = 0; | |
374 | goto out; | |
375 | } | |
94647322 DS |
376 | btrfs_err_rl(eb->fs_info, |
377 | "parent transid verify failed on %llu wanted %llu found %llu", | |
378 | eb->start, | |
29549aec | 379 | parent_transid, btrfs_header_generation(eb)); |
1259ab75 | 380 | ret = 1; |
a26e8c9f JB |
381 | |
382 | /* | |
383 | * Things reading via commit roots that don't have normal protection, | |
384 | * like send, can have a really old block in cache that may point at a | |
01327610 | 385 | * block that has been freed and re-allocated. So don't clear uptodate |
a26e8c9f JB |
386 | * if we find an eb that is under IO (dirty/writeback) because we could |
387 | * end up reading in the stale data and then writing it back out and | |
388 | * making everybody very sad. | |
389 | */ | |
390 | if (!extent_buffer_under_io(eb)) | |
391 | clear_extent_buffer_uptodate(eb); | |
33958dc6 | 392 | out: |
2ac55d41 JB |
393 | unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, |
394 | &cached_state, GFP_NOFS); | |
472b909f JB |
395 | if (need_lock) |
396 | btrfs_tree_read_unlock_blocking(eb); | |
1259ab75 | 397 | return ret; |
1259ab75 CM |
398 | } |
399 | ||
1104a885 DS |
400 | /* |
401 | * Return 0 if the superblock checksum type matches the checksum value of that | |
402 | * algorithm. Pass the raw disk superblock data. | |
403 | */ | |
ab8d0fc4 JM |
404 | static int btrfs_check_super_csum(struct btrfs_fs_info *fs_info, |
405 | char *raw_disk_sb) | |
1104a885 DS |
406 | { |
407 | struct btrfs_super_block *disk_sb = | |
408 | (struct btrfs_super_block *)raw_disk_sb; | |
409 | u16 csum_type = btrfs_super_csum_type(disk_sb); | |
410 | int ret = 0; | |
411 | ||
412 | if (csum_type == BTRFS_CSUM_TYPE_CRC32) { | |
413 | u32 crc = ~(u32)0; | |
414 | const int csum_size = sizeof(crc); | |
415 | char result[csum_size]; | |
416 | ||
417 | /* | |
418 | * The super_block structure does not span the whole | |
419 | * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space | |
01327610 | 420 | * is filled with zeros and is included in the checksum. |
1104a885 DS |
421 | */ |
422 | crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE, | |
423 | crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); | |
424 | btrfs_csum_final(crc, result); | |
425 | ||
426 | if (memcmp(raw_disk_sb, result, csum_size)) | |
427 | ret = 1; | |
428 | } | |
429 | ||
430 | if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) { | |
ab8d0fc4 | 431 | btrfs_err(fs_info, "unsupported checksum algorithm %u", |
1104a885 DS |
432 | csum_type); |
433 | ret = 1; | |
434 | } | |
435 | ||
436 | return ret; | |
437 | } | |
438 | ||
d352ac68 CM |
439 | /* |
440 | * helper to read a given tree block, doing retries as required when | |
441 | * the checksums don't match and we have alternate mirrors to try. | |
442 | */ | |
2ff7e61e | 443 | static int btree_read_extent_buffer_pages(struct btrfs_fs_info *fs_info, |
f188591e | 444 | struct extent_buffer *eb, |
8436ea91 | 445 | u64 parent_transid) |
f188591e CM |
446 | { |
447 | struct extent_io_tree *io_tree; | |
ea466794 | 448 | int failed = 0; |
f188591e CM |
449 | int ret; |
450 | int num_copies = 0; | |
451 | int mirror_num = 0; | |
ea466794 | 452 | int failed_mirror = 0; |
f188591e | 453 | |
a826d6dc | 454 | clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
0b246afa | 455 | io_tree = &BTRFS_I(fs_info->btree_inode)->io_tree; |
f188591e | 456 | while (1) { |
8436ea91 | 457 | ret = read_extent_buffer_pages(io_tree, eb, WAIT_COMPLETE, |
f188591e | 458 | btree_get_extent, mirror_num); |
256dd1bb SB |
459 | if (!ret) { |
460 | if (!verify_parent_transid(io_tree, eb, | |
b9fab919 | 461 | parent_transid, 0)) |
256dd1bb SB |
462 | break; |
463 | else | |
464 | ret = -EIO; | |
465 | } | |
d397712b | 466 | |
a826d6dc JB |
467 | /* |
468 | * This buffer's crc is fine, but its contents are corrupted, so | |
469 | * there is no reason to read the other copies, they won't be | |
470 | * any less wrong. | |
471 | */ | |
472 | if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags)) | |
ea466794 JB |
473 | break; |
474 | ||
0b246afa | 475 | num_copies = btrfs_num_copies(fs_info, |
f188591e | 476 | eb->start, eb->len); |
4235298e | 477 | if (num_copies == 1) |
ea466794 | 478 | break; |
4235298e | 479 | |
5cf1ab56 JB |
480 | if (!failed_mirror) { |
481 | failed = 1; | |
482 | failed_mirror = eb->read_mirror; | |
483 | } | |
484 | ||
f188591e | 485 | mirror_num++; |
ea466794 JB |
486 | if (mirror_num == failed_mirror) |
487 | mirror_num++; | |
488 | ||
4235298e | 489 | if (mirror_num > num_copies) |
ea466794 | 490 | break; |
f188591e | 491 | } |
ea466794 | 492 | |
c0901581 | 493 | if (failed && !ret && failed_mirror) |
2ff7e61e | 494 | repair_eb_io_failure(fs_info, eb, failed_mirror); |
ea466794 JB |
495 | |
496 | return ret; | |
f188591e | 497 | } |
19c00ddc | 498 | |
d352ac68 | 499 | /* |
d397712b CM |
500 | * checksum a dirty tree block before IO. This has extra checks to make sure |
501 | * we only fill in the checksum field in the first page of a multi-page block | |
d352ac68 | 502 | */ |
d397712b | 503 | |
01d58472 | 504 | static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct page *page) |
19c00ddc | 505 | { |
4eee4fa4 | 506 | u64 start = page_offset(page); |
19c00ddc | 507 | u64 found_start; |
19c00ddc | 508 | struct extent_buffer *eb; |
f188591e | 509 | |
4f2de97a JB |
510 | eb = (struct extent_buffer *)page->private; |
511 | if (page != eb->pages[0]) | |
512 | return 0; | |
0f805531 | 513 | |
19c00ddc | 514 | found_start = btrfs_header_bytenr(eb); |
0f805531 AL |
515 | /* |
516 | * Please do not consolidate these warnings into a single if. | |
517 | * It is useful to know what went wrong. | |
518 | */ | |
519 | if (WARN_ON(found_start != start)) | |
520 | return -EUCLEAN; | |
521 | if (WARN_ON(!PageUptodate(page))) | |
522 | return -EUCLEAN; | |
523 | ||
524 | ASSERT(memcmp_extent_buffer(eb, fs_info->fsid, | |
525 | btrfs_header_fsid(), BTRFS_FSID_SIZE) == 0); | |
526 | ||
8bd98f0e | 527 | return csum_tree_block(fs_info, eb, 0); |
19c00ddc CM |
528 | } |
529 | ||
01d58472 | 530 | static int check_tree_block_fsid(struct btrfs_fs_info *fs_info, |
2b82032c YZ |
531 | struct extent_buffer *eb) |
532 | { | |
01d58472 | 533 | struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; |
44880fdc | 534 | u8 fsid[BTRFS_FSID_SIZE]; |
2b82032c YZ |
535 | int ret = 1; |
536 | ||
0a4e5586 | 537 | read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE); |
2b82032c YZ |
538 | while (fs_devices) { |
539 | if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { | |
540 | ret = 0; | |
541 | break; | |
542 | } | |
543 | fs_devices = fs_devices->seed; | |
544 | } | |
545 | return ret; | |
546 | } | |
547 | ||
facc8a22 MX |
548 | static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio, |
549 | u64 phy_offset, struct page *page, | |
550 | u64 start, u64 end, int mirror) | |
ce9adaa5 | 551 | { |
ce9adaa5 CM |
552 | u64 found_start; |
553 | int found_level; | |
ce9adaa5 CM |
554 | struct extent_buffer *eb; |
555 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
02873e43 | 556 | struct btrfs_fs_info *fs_info = root->fs_info; |
f188591e | 557 | int ret = 0; |
727011e0 | 558 | int reads_done; |
ce9adaa5 | 559 | |
ce9adaa5 CM |
560 | if (!page->private) |
561 | goto out; | |
d397712b | 562 | |
4f2de97a | 563 | eb = (struct extent_buffer *)page->private; |
d397712b | 564 | |
0b32f4bb JB |
565 | /* the pending IO might have been the only thing that kept this buffer |
566 | * in memory. Make sure we have a ref for all this other checks | |
567 | */ | |
568 | extent_buffer_get(eb); | |
569 | ||
570 | reads_done = atomic_dec_and_test(&eb->io_pages); | |
727011e0 CM |
571 | if (!reads_done) |
572 | goto err; | |
f188591e | 573 | |
5cf1ab56 | 574 | eb->read_mirror = mirror; |
656f30db | 575 | if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { |
ea466794 JB |
576 | ret = -EIO; |
577 | goto err; | |
578 | } | |
579 | ||
ce9adaa5 | 580 | found_start = btrfs_header_bytenr(eb); |
727011e0 | 581 | if (found_start != eb->start) { |
02873e43 ZL |
582 | btrfs_err_rl(fs_info, "bad tree block start %llu %llu", |
583 | found_start, eb->start); | |
f188591e | 584 | ret = -EIO; |
ce9adaa5 CM |
585 | goto err; |
586 | } | |
02873e43 ZL |
587 | if (check_tree_block_fsid(fs_info, eb)) { |
588 | btrfs_err_rl(fs_info, "bad fsid on block %llu", | |
589 | eb->start); | |
1259ab75 CM |
590 | ret = -EIO; |
591 | goto err; | |
592 | } | |
ce9adaa5 | 593 | found_level = btrfs_header_level(eb); |
1c24c3ce | 594 | if (found_level >= BTRFS_MAX_LEVEL) { |
02873e43 ZL |
595 | btrfs_err(fs_info, "bad tree block level %d", |
596 | (int)btrfs_header_level(eb)); | |
1c24c3ce JB |
597 | ret = -EIO; |
598 | goto err; | |
599 | } | |
ce9adaa5 | 600 | |
85d4e461 CM |
601 | btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), |
602 | eb, found_level); | |
4008c04a | 603 | |
02873e43 | 604 | ret = csum_tree_block(fs_info, eb, 1); |
8bd98f0e | 605 | if (ret) |
a826d6dc | 606 | goto err; |
a826d6dc JB |
607 | |
608 | /* | |
609 | * If this is a leaf block and it is corrupt, set the corrupt bit so | |
610 | * that we don't try and read the other copies of this block, just | |
611 | * return -EIO. | |
612 | */ | |
557ea5dd | 613 | if (found_level == 0 && btrfs_check_leaf(root, eb)) { |
a826d6dc JB |
614 | set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
615 | ret = -EIO; | |
616 | } | |
ce9adaa5 | 617 | |
557ea5dd | 618 | if (found_level > 0 && btrfs_check_node(root, eb)) |
053ab70f LB |
619 | ret = -EIO; |
620 | ||
0b32f4bb JB |
621 | if (!ret) |
622 | set_extent_buffer_uptodate(eb); | |
ce9adaa5 | 623 | err: |
79fb65a1 JB |
624 | if (reads_done && |
625 | test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) | |
d48d71aa | 626 | btree_readahead_hook(eb, ret); |
4bb31e92 | 627 | |
53b381b3 DW |
628 | if (ret) { |
629 | /* | |
630 | * our io error hook is going to dec the io pages | |
631 | * again, we have to make sure it has something | |
632 | * to decrement | |
633 | */ | |
634 | atomic_inc(&eb->io_pages); | |
0b32f4bb | 635 | clear_extent_buffer_uptodate(eb); |
53b381b3 | 636 | } |
0b32f4bb | 637 | free_extent_buffer(eb); |
ce9adaa5 | 638 | out: |
f188591e | 639 | return ret; |
ce9adaa5 CM |
640 | } |
641 | ||
ea466794 | 642 | static int btree_io_failed_hook(struct page *page, int failed_mirror) |
4bb31e92 | 643 | { |
4bb31e92 | 644 | struct extent_buffer *eb; |
4bb31e92 | 645 | |
4f2de97a | 646 | eb = (struct extent_buffer *)page->private; |
656f30db | 647 | set_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags); |
5cf1ab56 | 648 | eb->read_mirror = failed_mirror; |
53b381b3 | 649 | atomic_dec(&eb->io_pages); |
ea466794 | 650 | if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) |
d48d71aa | 651 | btree_readahead_hook(eb, -EIO); |
4bb31e92 AJ |
652 | return -EIO; /* we fixed nothing */ |
653 | } | |
654 | ||
4246a0b6 | 655 | static void end_workqueue_bio(struct bio *bio) |
ce9adaa5 | 656 | { |
97eb6b69 | 657 | struct btrfs_end_io_wq *end_io_wq = bio->bi_private; |
ce9adaa5 | 658 | struct btrfs_fs_info *fs_info; |
9e0af237 LB |
659 | struct btrfs_workqueue *wq; |
660 | btrfs_work_func_t func; | |
ce9adaa5 | 661 | |
ce9adaa5 | 662 | fs_info = end_io_wq->info; |
4e4cbee9 | 663 | end_io_wq->status = bio->bi_status; |
d20f7043 | 664 | |
37226b21 | 665 | if (bio_op(bio) == REQ_OP_WRITE) { |
9e0af237 LB |
666 | if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) { |
667 | wq = fs_info->endio_meta_write_workers; | |
668 | func = btrfs_endio_meta_write_helper; | |
669 | } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) { | |
670 | wq = fs_info->endio_freespace_worker; | |
671 | func = btrfs_freespace_write_helper; | |
672 | } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) { | |
673 | wq = fs_info->endio_raid56_workers; | |
674 | func = btrfs_endio_raid56_helper; | |
675 | } else { | |
676 | wq = fs_info->endio_write_workers; | |
677 | func = btrfs_endio_write_helper; | |
678 | } | |
d20f7043 | 679 | } else { |
8b110e39 MX |
680 | if (unlikely(end_io_wq->metadata == |
681 | BTRFS_WQ_ENDIO_DIO_REPAIR)) { | |
682 | wq = fs_info->endio_repair_workers; | |
683 | func = btrfs_endio_repair_helper; | |
684 | } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) { | |
9e0af237 LB |
685 | wq = fs_info->endio_raid56_workers; |
686 | func = btrfs_endio_raid56_helper; | |
687 | } else if (end_io_wq->metadata) { | |
688 | wq = fs_info->endio_meta_workers; | |
689 | func = btrfs_endio_meta_helper; | |
690 | } else { | |
691 | wq = fs_info->endio_workers; | |
692 | func = btrfs_endio_helper; | |
693 | } | |
d20f7043 | 694 | } |
9e0af237 LB |
695 | |
696 | btrfs_init_work(&end_io_wq->work, func, end_workqueue_fn, NULL, NULL); | |
697 | btrfs_queue_work(wq, &end_io_wq->work); | |
ce9adaa5 CM |
698 | } |
699 | ||
4e4cbee9 | 700 | blk_status_t btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
bfebd8b5 | 701 | enum btrfs_wq_endio_type metadata) |
0b86a832 | 702 | { |
97eb6b69 | 703 | struct btrfs_end_io_wq *end_io_wq; |
8b110e39 | 704 | |
97eb6b69 | 705 | end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS); |
ce9adaa5 | 706 | if (!end_io_wq) |
4e4cbee9 | 707 | return BLK_STS_RESOURCE; |
ce9adaa5 CM |
708 | |
709 | end_io_wq->private = bio->bi_private; | |
710 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 711 | end_io_wq->info = info; |
4e4cbee9 | 712 | end_io_wq->status = 0; |
ce9adaa5 | 713 | end_io_wq->bio = bio; |
22c59948 | 714 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
715 | |
716 | bio->bi_private = end_io_wq; | |
717 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
718 | return 0; |
719 | } | |
720 | ||
b64a2851 | 721 | unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) |
0986fe9e | 722 | { |
4854ddd0 | 723 | unsigned long limit = min_t(unsigned long, |
5cdc7ad3 | 724 | info->thread_pool_size, |
4854ddd0 CM |
725 | info->fs_devices->open_devices); |
726 | return 256 * limit; | |
727 | } | |
0986fe9e | 728 | |
4a69a410 CM |
729 | static void run_one_async_start(struct btrfs_work *work) |
730 | { | |
4a69a410 | 731 | struct async_submit_bio *async; |
4e4cbee9 | 732 | blk_status_t ret; |
4a69a410 CM |
733 | |
734 | async = container_of(work, struct async_submit_bio, work); | |
c6100a4b | 735 | ret = async->submit_bio_start(async->private_data, async->bio, |
79787eaa JM |
736 | async->mirror_num, async->bio_flags, |
737 | async->bio_offset); | |
738 | if (ret) | |
4e4cbee9 | 739 | async->status = ret; |
4a69a410 CM |
740 | } |
741 | ||
742 | static void run_one_async_done(struct btrfs_work *work) | |
8b712842 | 743 | { |
8b712842 CM |
744 | struct async_submit_bio *async; |
745 | ||
746 | async = container_of(work, struct async_submit_bio, work); | |
4854ddd0 | 747 | |
bb7ab3b9 | 748 | /* If an error occurred we just want to clean up the bio and move on */ |
4e4cbee9 CH |
749 | if (async->status) { |
750 | async->bio->bi_status = async->status; | |
4246a0b6 | 751 | bio_endio(async->bio); |
79787eaa JM |
752 | return; |
753 | } | |
754 | ||
c6100a4b | 755 | async->submit_bio_done(async->private_data, async->bio, async->mirror_num, |
81a75f67 | 756 | async->bio_flags, async->bio_offset); |
4a69a410 CM |
757 | } |
758 | ||
759 | static void run_one_async_free(struct btrfs_work *work) | |
760 | { | |
761 | struct async_submit_bio *async; | |
762 | ||
763 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
764 | kfree(async); |
765 | } | |
766 | ||
8c27cb35 LT |
767 | blk_status_t btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct bio *bio, |
768 | int mirror_num, unsigned long bio_flags, | |
769 | u64 bio_offset, void *private_data, | |
770 | extent_submit_bio_hook_t *submit_bio_start, | |
771 | extent_submit_bio_hook_t *submit_bio_done) | |
44b8bd7e CM |
772 | { |
773 | struct async_submit_bio *async; | |
774 | ||
775 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
776 | if (!async) | |
4e4cbee9 | 777 | return BLK_STS_RESOURCE; |
44b8bd7e | 778 | |
c6100a4b JB |
779 | async->private_data = private_data; |
780 | async->fs_info = fs_info; | |
44b8bd7e CM |
781 | async->bio = bio; |
782 | async->mirror_num = mirror_num; | |
4a69a410 CM |
783 | async->submit_bio_start = submit_bio_start; |
784 | async->submit_bio_done = submit_bio_done; | |
785 | ||
9e0af237 | 786 | btrfs_init_work(&async->work, btrfs_worker_helper, run_one_async_start, |
5cdc7ad3 | 787 | run_one_async_done, run_one_async_free); |
4a69a410 | 788 | |
c8b97818 | 789 | async->bio_flags = bio_flags; |
eaf25d93 | 790 | async->bio_offset = bio_offset; |
8c8bee1d | 791 | |
4e4cbee9 | 792 | async->status = 0; |
79787eaa | 793 | |
67f055c7 | 794 | if (op_is_sync(bio->bi_opf)) |
5cdc7ad3 | 795 | btrfs_set_work_high_priority(&async->work); |
d313d7a3 | 796 | |
5cdc7ad3 | 797 | btrfs_queue_work(fs_info->workers, &async->work); |
44b8bd7e CM |
798 | return 0; |
799 | } | |
800 | ||
4e4cbee9 | 801 | static blk_status_t btree_csum_one_bio(struct bio *bio) |
ce3ed71a | 802 | { |
2c30c71b | 803 | struct bio_vec *bvec; |
ce3ed71a | 804 | struct btrfs_root *root; |
2c30c71b | 805 | int i, ret = 0; |
ce3ed71a | 806 | |
c09abff8 | 807 | ASSERT(!bio_flagged(bio, BIO_CLONED)); |
2c30c71b | 808 | bio_for_each_segment_all(bvec, bio, i) { |
ce3ed71a | 809 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
01d58472 | 810 | ret = csum_dirty_buffer(root->fs_info, bvec->bv_page); |
79787eaa JM |
811 | if (ret) |
812 | break; | |
ce3ed71a | 813 | } |
2c30c71b | 814 | |
4e4cbee9 | 815 | return errno_to_blk_status(ret); |
ce3ed71a CM |
816 | } |
817 | ||
8c27cb35 LT |
818 | static blk_status_t __btree_submit_bio_start(void *private_data, struct bio *bio, |
819 | int mirror_num, unsigned long bio_flags, | |
820 | u64 bio_offset) | |
22c59948 | 821 | { |
8b712842 CM |
822 | /* |
823 | * when we're called for a write, we're already in the async | |
5443be45 | 824 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 825 | */ |
79787eaa | 826 | return btree_csum_one_bio(bio); |
4a69a410 | 827 | } |
22c59948 | 828 | |
8c27cb35 LT |
829 | static blk_status_t __btree_submit_bio_done(void *private_data, struct bio *bio, |
830 | int mirror_num, unsigned long bio_flags, | |
831 | u64 bio_offset) | |
4a69a410 | 832 | { |
c6100a4b | 833 | struct inode *inode = private_data; |
4e4cbee9 | 834 | blk_status_t ret; |
61891923 | 835 | |
8b712842 | 836 | /* |
4a69a410 CM |
837 | * when we're called for a write, we're already in the async |
838 | * submission context. Just jump into btrfs_map_bio | |
8b712842 | 839 | */ |
2ff7e61e | 840 | ret = btrfs_map_bio(btrfs_sb(inode->i_sb), bio, mirror_num, 1); |
4246a0b6 | 841 | if (ret) { |
4e4cbee9 | 842 | bio->bi_status = ret; |
4246a0b6 CH |
843 | bio_endio(bio); |
844 | } | |
61891923 | 845 | return ret; |
0b86a832 CM |
846 | } |
847 | ||
18fdc679 | 848 | static int check_async_write(struct btrfs_inode *bi) |
de0022b9 | 849 | { |
6300463b LB |
850 | if (atomic_read(&bi->sync_writers)) |
851 | return 0; | |
de0022b9 | 852 | #ifdef CONFIG_X86 |
bc696ca0 | 853 | if (static_cpu_has(X86_FEATURE_XMM4_2)) |
de0022b9 JB |
854 | return 0; |
855 | #endif | |
856 | return 1; | |
857 | } | |
858 | ||
8c27cb35 LT |
859 | static blk_status_t btree_submit_bio_hook(void *private_data, struct bio *bio, |
860 | int mirror_num, unsigned long bio_flags, | |
861 | u64 bio_offset) | |
44b8bd7e | 862 | { |
c6100a4b | 863 | struct inode *inode = private_data; |
0b246afa | 864 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
18fdc679 | 865 | int async = check_async_write(BTRFS_I(inode)); |
4e4cbee9 | 866 | blk_status_t ret; |
cad321ad | 867 | |
37226b21 | 868 | if (bio_op(bio) != REQ_OP_WRITE) { |
4a69a410 CM |
869 | /* |
870 | * called for a read, do the setup so that checksum validation | |
871 | * can happen in the async kernel threads | |
872 | */ | |
0b246afa JM |
873 | ret = btrfs_bio_wq_end_io(fs_info, bio, |
874 | BTRFS_WQ_ENDIO_METADATA); | |
1d4284bd | 875 | if (ret) |
61891923 | 876 | goto out_w_error; |
2ff7e61e | 877 | ret = btrfs_map_bio(fs_info, bio, mirror_num, 0); |
de0022b9 JB |
878 | } else if (!async) { |
879 | ret = btree_csum_one_bio(bio); | |
880 | if (ret) | |
61891923 | 881 | goto out_w_error; |
2ff7e61e | 882 | ret = btrfs_map_bio(fs_info, bio, mirror_num, 0); |
61891923 SB |
883 | } else { |
884 | /* | |
885 | * kthread helpers are used to submit writes so that | |
886 | * checksumming can happen in parallel across all CPUs | |
887 | */ | |
c6100a4b JB |
888 | ret = btrfs_wq_submit_bio(fs_info, bio, mirror_num, 0, |
889 | bio_offset, private_data, | |
61891923 SB |
890 | __btree_submit_bio_start, |
891 | __btree_submit_bio_done); | |
44b8bd7e | 892 | } |
d313d7a3 | 893 | |
4246a0b6 CH |
894 | if (ret) |
895 | goto out_w_error; | |
896 | return 0; | |
897 | ||
61891923 | 898 | out_w_error: |
4e4cbee9 | 899 | bio->bi_status = ret; |
4246a0b6 | 900 | bio_endio(bio); |
61891923 | 901 | return ret; |
44b8bd7e CM |
902 | } |
903 | ||
3dd1462e | 904 | #ifdef CONFIG_MIGRATION |
784b4e29 | 905 | static int btree_migratepage(struct address_space *mapping, |
a6bc32b8 MG |
906 | struct page *newpage, struct page *page, |
907 | enum migrate_mode mode) | |
784b4e29 CM |
908 | { |
909 | /* | |
910 | * we can't safely write a btree page from here, | |
911 | * we haven't done the locking hook | |
912 | */ | |
913 | if (PageDirty(page)) | |
914 | return -EAGAIN; | |
915 | /* | |
916 | * Buffers may be managed in a filesystem specific way. | |
917 | * We must have no buffers or drop them. | |
918 | */ | |
919 | if (page_has_private(page) && | |
920 | !try_to_release_page(page, GFP_KERNEL)) | |
921 | return -EAGAIN; | |
a6bc32b8 | 922 | return migrate_page(mapping, newpage, page, mode); |
784b4e29 | 923 | } |
3dd1462e | 924 | #endif |
784b4e29 | 925 | |
0da5468f CM |
926 | |
927 | static int btree_writepages(struct address_space *mapping, | |
928 | struct writeback_control *wbc) | |
929 | { | |
e2d84521 MX |
930 | struct btrfs_fs_info *fs_info; |
931 | int ret; | |
932 | ||
d8d5f3e1 | 933 | if (wbc->sync_mode == WB_SYNC_NONE) { |
448d640b CM |
934 | |
935 | if (wbc->for_kupdate) | |
936 | return 0; | |
937 | ||
e2d84521 | 938 | fs_info = BTRFS_I(mapping->host)->root->fs_info; |
b9473439 | 939 | /* this is a bit racy, but that's ok */ |
e2d84521 MX |
940 | ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
941 | BTRFS_DIRTY_METADATA_THRESH); | |
942 | if (ret < 0) | |
793955bc | 943 | return 0; |
793955bc | 944 | } |
0b32f4bb | 945 | return btree_write_cache_pages(mapping, wbc); |
0da5468f CM |
946 | } |
947 | ||
b2950863 | 948 | static int btree_readpage(struct file *file, struct page *page) |
5f39d397 | 949 | { |
d1310b2e CM |
950 | struct extent_io_tree *tree; |
951 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
8ddc7d9c | 952 | return extent_read_full_page(tree, page, btree_get_extent, 0); |
5f39d397 | 953 | } |
22b0ebda | 954 | |
70dec807 | 955 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 956 | { |
98509cfc | 957 | if (PageWriteback(page) || PageDirty(page)) |
d397712b | 958 | return 0; |
0c4e538b | 959 | |
f7a52a40 | 960 | return try_release_extent_buffer(page); |
d98237b3 CM |
961 | } |
962 | ||
d47992f8 LC |
963 | static void btree_invalidatepage(struct page *page, unsigned int offset, |
964 | unsigned int length) | |
d98237b3 | 965 | { |
d1310b2e CM |
966 | struct extent_io_tree *tree; |
967 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
968 | extent_invalidatepage(tree, page, offset); |
969 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 970 | if (PagePrivate(page)) { |
efe120a0 FH |
971 | btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info, |
972 | "page private not zero on page %llu", | |
973 | (unsigned long long)page_offset(page)); | |
9ad6b7bc CM |
974 | ClearPagePrivate(page); |
975 | set_page_private(page, 0); | |
09cbfeaf | 976 | put_page(page); |
9ad6b7bc | 977 | } |
d98237b3 CM |
978 | } |
979 | ||
0b32f4bb JB |
980 | static int btree_set_page_dirty(struct page *page) |
981 | { | |
bb146eb2 | 982 | #ifdef DEBUG |
0b32f4bb JB |
983 | struct extent_buffer *eb; |
984 | ||
985 | BUG_ON(!PagePrivate(page)); | |
986 | eb = (struct extent_buffer *)page->private; | |
987 | BUG_ON(!eb); | |
988 | BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
989 | BUG_ON(!atomic_read(&eb->refs)); | |
990 | btrfs_assert_tree_locked(eb); | |
bb146eb2 | 991 | #endif |
0b32f4bb JB |
992 | return __set_page_dirty_nobuffers(page); |
993 | } | |
994 | ||
7f09410b | 995 | static const struct address_space_operations btree_aops = { |
d98237b3 | 996 | .readpage = btree_readpage, |
0da5468f | 997 | .writepages = btree_writepages, |
5f39d397 CM |
998 | .releasepage = btree_releasepage, |
999 | .invalidatepage = btree_invalidatepage, | |
5a92bc88 | 1000 | #ifdef CONFIG_MIGRATION |
784b4e29 | 1001 | .migratepage = btree_migratepage, |
5a92bc88 | 1002 | #endif |
0b32f4bb | 1003 | .set_page_dirty = btree_set_page_dirty, |
d98237b3 CM |
1004 | }; |
1005 | ||
2ff7e61e | 1006 | void readahead_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr) |
090d1875 | 1007 | { |
5f39d397 | 1008 | struct extent_buffer *buf = NULL; |
2ff7e61e | 1009 | struct inode *btree_inode = fs_info->btree_inode; |
090d1875 | 1010 | |
2ff7e61e | 1011 | buf = btrfs_find_create_tree_block(fs_info, bytenr); |
c871b0f2 | 1012 | if (IS_ERR(buf)) |
6197d86e | 1013 | return; |
d1310b2e | 1014 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, |
8436ea91 | 1015 | buf, WAIT_NONE, btree_get_extent, 0); |
5f39d397 | 1016 | free_extent_buffer(buf); |
090d1875 CM |
1017 | } |
1018 | ||
2ff7e61e | 1019 | int reada_tree_block_flagged(struct btrfs_fs_info *fs_info, u64 bytenr, |
ab0fff03 AJ |
1020 | int mirror_num, struct extent_buffer **eb) |
1021 | { | |
1022 | struct extent_buffer *buf = NULL; | |
2ff7e61e | 1023 | struct inode *btree_inode = fs_info->btree_inode; |
ab0fff03 AJ |
1024 | struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree; |
1025 | int ret; | |
1026 | ||
2ff7e61e | 1027 | buf = btrfs_find_create_tree_block(fs_info, bytenr); |
c871b0f2 | 1028 | if (IS_ERR(buf)) |
ab0fff03 AJ |
1029 | return 0; |
1030 | ||
1031 | set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags); | |
1032 | ||
8436ea91 | 1033 | ret = read_extent_buffer_pages(io_tree, buf, WAIT_PAGE_LOCK, |
ab0fff03 AJ |
1034 | btree_get_extent, mirror_num); |
1035 | if (ret) { | |
1036 | free_extent_buffer(buf); | |
1037 | return ret; | |
1038 | } | |
1039 | ||
1040 | if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) { | |
1041 | free_extent_buffer(buf); | |
1042 | return -EIO; | |
0b32f4bb | 1043 | } else if (extent_buffer_uptodate(buf)) { |
ab0fff03 AJ |
1044 | *eb = buf; |
1045 | } else { | |
1046 | free_extent_buffer(buf); | |
1047 | } | |
1048 | return 0; | |
1049 | } | |
1050 | ||
2ff7e61e JM |
1051 | struct extent_buffer *btrfs_find_create_tree_block( |
1052 | struct btrfs_fs_info *fs_info, | |
1053 | u64 bytenr) | |
0999df54 | 1054 | { |
0b246afa JM |
1055 | if (btrfs_is_testing(fs_info)) |
1056 | return alloc_test_extent_buffer(fs_info, bytenr); | |
1057 | return alloc_extent_buffer(fs_info, bytenr); | |
0999df54 CM |
1058 | } |
1059 | ||
1060 | ||
e02119d5 CM |
1061 | int btrfs_write_tree_block(struct extent_buffer *buf) |
1062 | { | |
727011e0 | 1063 | return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start, |
8aa38c31 | 1064 | buf->start + buf->len - 1); |
e02119d5 CM |
1065 | } |
1066 | ||
3189ff77 | 1067 | void btrfs_wait_tree_block_writeback(struct extent_buffer *buf) |
e02119d5 | 1068 | { |
3189ff77 JL |
1069 | filemap_fdatawait_range(buf->pages[0]->mapping, |
1070 | buf->start, buf->start + buf->len - 1); | |
e02119d5 CM |
1071 | } |
1072 | ||
2ff7e61e | 1073 | struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, |
ce86cd59 | 1074 | u64 parent_transid) |
0999df54 CM |
1075 | { |
1076 | struct extent_buffer *buf = NULL; | |
0999df54 CM |
1077 | int ret; |
1078 | ||
2ff7e61e | 1079 | buf = btrfs_find_create_tree_block(fs_info, bytenr); |
c871b0f2 LB |
1080 | if (IS_ERR(buf)) |
1081 | return buf; | |
0999df54 | 1082 | |
2ff7e61e | 1083 | ret = btree_read_extent_buffer_pages(fs_info, buf, parent_transid); |
0f0fe8f7 FDBM |
1084 | if (ret) { |
1085 | free_extent_buffer(buf); | |
64c043de | 1086 | return ERR_PTR(ret); |
0f0fe8f7 | 1087 | } |
5f39d397 | 1088 | return buf; |
ce9adaa5 | 1089 | |
eb60ceac CM |
1090 | } |
1091 | ||
7c302b49 | 1092 | void clean_tree_block(struct btrfs_fs_info *fs_info, |
d5c13f92 | 1093 | struct extent_buffer *buf) |
ed2ff2cb | 1094 | { |
55c69072 | 1095 | if (btrfs_header_generation(buf) == |
e2d84521 | 1096 | fs_info->running_transaction->transid) { |
b9447ef8 | 1097 | btrfs_assert_tree_locked(buf); |
b4ce94de | 1098 | |
b9473439 | 1099 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { |
104b4e51 NB |
1100 | percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, |
1101 | -buf->len, | |
1102 | fs_info->dirty_metadata_batch); | |
ed7b63eb JB |
1103 | /* ugh, clear_extent_buffer_dirty needs to lock the page */ |
1104 | btrfs_set_lock_blocking(buf); | |
1105 | clear_extent_buffer_dirty(buf); | |
1106 | } | |
925baedd | 1107 | } |
5f39d397 CM |
1108 | } |
1109 | ||
8257b2dc MX |
1110 | static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void) |
1111 | { | |
1112 | struct btrfs_subvolume_writers *writers; | |
1113 | int ret; | |
1114 | ||
1115 | writers = kmalloc(sizeof(*writers), GFP_NOFS); | |
1116 | if (!writers) | |
1117 | return ERR_PTR(-ENOMEM); | |
1118 | ||
908c7f19 | 1119 | ret = percpu_counter_init(&writers->counter, 0, GFP_KERNEL); |
8257b2dc MX |
1120 | if (ret < 0) { |
1121 | kfree(writers); | |
1122 | return ERR_PTR(ret); | |
1123 | } | |
1124 | ||
1125 | init_waitqueue_head(&writers->wait); | |
1126 | return writers; | |
1127 | } | |
1128 | ||
1129 | static void | |
1130 | btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers) | |
1131 | { | |
1132 | percpu_counter_destroy(&writers->counter); | |
1133 | kfree(writers); | |
1134 | } | |
1135 | ||
da17066c | 1136 | static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, |
143bede5 | 1137 | u64 objectid) |
d97e63b6 | 1138 | { |
7c0260ee | 1139 | bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); |
cfaa7295 | 1140 | root->node = NULL; |
a28ec197 | 1141 | root->commit_root = NULL; |
27cdeb70 | 1142 | root->state = 0; |
d68fc57b | 1143 | root->orphan_cleanup_state = 0; |
0b86a832 | 1144 | |
0f7d52f4 CM |
1145 | root->objectid = objectid; |
1146 | root->last_trans = 0; | |
13a8a7c8 | 1147 | root->highest_objectid = 0; |
eb73c1b7 | 1148 | root->nr_delalloc_inodes = 0; |
199c2a9c | 1149 | root->nr_ordered_extents = 0; |
58176a96 | 1150 | root->name = NULL; |
6bef4d31 | 1151 | root->inode_tree = RB_ROOT; |
16cdcec7 | 1152 | INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); |
f0486c68 | 1153 | root->block_rsv = NULL; |
d68fc57b | 1154 | root->orphan_block_rsv = NULL; |
0b86a832 CM |
1155 | |
1156 | INIT_LIST_HEAD(&root->dirty_list); | |
5d4f98a2 | 1157 | INIT_LIST_HEAD(&root->root_list); |
eb73c1b7 MX |
1158 | INIT_LIST_HEAD(&root->delalloc_inodes); |
1159 | INIT_LIST_HEAD(&root->delalloc_root); | |
199c2a9c MX |
1160 | INIT_LIST_HEAD(&root->ordered_extents); |
1161 | INIT_LIST_HEAD(&root->ordered_root); | |
2ab28f32 JB |
1162 | INIT_LIST_HEAD(&root->logged_list[0]); |
1163 | INIT_LIST_HEAD(&root->logged_list[1]); | |
d68fc57b | 1164 | spin_lock_init(&root->orphan_lock); |
5d4f98a2 | 1165 | spin_lock_init(&root->inode_lock); |
eb73c1b7 | 1166 | spin_lock_init(&root->delalloc_lock); |
199c2a9c | 1167 | spin_lock_init(&root->ordered_extent_lock); |
f0486c68 | 1168 | spin_lock_init(&root->accounting_lock); |
2ab28f32 JB |
1169 | spin_lock_init(&root->log_extents_lock[0]); |
1170 | spin_lock_init(&root->log_extents_lock[1]); | |
a2135011 | 1171 | mutex_init(&root->objectid_mutex); |
e02119d5 | 1172 | mutex_init(&root->log_mutex); |
31f3d255 | 1173 | mutex_init(&root->ordered_extent_mutex); |
573bfb72 | 1174 | mutex_init(&root->delalloc_mutex); |
7237f183 YZ |
1175 | init_waitqueue_head(&root->log_writer_wait); |
1176 | init_waitqueue_head(&root->log_commit_wait[0]); | |
1177 | init_waitqueue_head(&root->log_commit_wait[1]); | |
8b050d35 MX |
1178 | INIT_LIST_HEAD(&root->log_ctxs[0]); |
1179 | INIT_LIST_HEAD(&root->log_ctxs[1]); | |
7237f183 YZ |
1180 | atomic_set(&root->log_commit[0], 0); |
1181 | atomic_set(&root->log_commit[1], 0); | |
1182 | atomic_set(&root->log_writers, 0); | |
2ecb7923 | 1183 | atomic_set(&root->log_batch, 0); |
8a35d95f | 1184 | atomic_set(&root->orphan_inodes, 0); |
0700cea7 | 1185 | refcount_set(&root->refs, 1); |
ea14b57f | 1186 | atomic_set(&root->will_be_snapshotted, 0); |
ce0dcee6 | 1187 | atomic64_set(&root->qgroup_meta_rsv, 0); |
7237f183 | 1188 | root->log_transid = 0; |
d1433deb | 1189 | root->log_transid_committed = -1; |
257c62e1 | 1190 | root->last_log_commit = 0; |
7c0260ee | 1191 | if (!dummy) |
c6100a4b | 1192 | extent_io_tree_init(&root->dirty_log_pages, NULL); |
017e5369 | 1193 | |
3768f368 CM |
1194 | memset(&root->root_key, 0, sizeof(root->root_key)); |
1195 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 1196 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
7c0260ee | 1197 | if (!dummy) |
06ea65a3 JB |
1198 | root->defrag_trans_start = fs_info->generation; |
1199 | else | |
1200 | root->defrag_trans_start = 0; | |
4d775673 | 1201 | root->root_key.objectid = objectid; |
0ee5dc67 | 1202 | root->anon_dev = 0; |
8ea05e3a | 1203 | |
5f3ab90a | 1204 | spin_lock_init(&root->root_item_lock); |
3768f368 CM |
1205 | } |
1206 | ||
74e4d827 DS |
1207 | static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, |
1208 | gfp_t flags) | |
6f07e42e | 1209 | { |
74e4d827 | 1210 | struct btrfs_root *root = kzalloc(sizeof(*root), flags); |
6f07e42e AV |
1211 | if (root) |
1212 | root->fs_info = fs_info; | |
1213 | return root; | |
1214 | } | |
1215 | ||
06ea65a3 JB |
1216 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
1217 | /* Should only be used by the testing infrastructure */ | |
da17066c | 1218 | struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) |
06ea65a3 JB |
1219 | { |
1220 | struct btrfs_root *root; | |
1221 | ||
7c0260ee JM |
1222 | if (!fs_info) |
1223 | return ERR_PTR(-EINVAL); | |
1224 | ||
1225 | root = btrfs_alloc_root(fs_info, GFP_KERNEL); | |
06ea65a3 JB |
1226 | if (!root) |
1227 | return ERR_PTR(-ENOMEM); | |
da17066c | 1228 | |
b9ef22de | 1229 | /* We don't use the stripesize in selftest, set it as sectorsize */ |
da17066c | 1230 | __setup_root(root, fs_info, BTRFS_ROOT_TREE_OBJECTID); |
faa2dbf0 | 1231 | root->alloc_bytenr = 0; |
06ea65a3 JB |
1232 | |
1233 | return root; | |
1234 | } | |
1235 | #endif | |
1236 | ||
20897f5c AJ |
1237 | struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, |
1238 | struct btrfs_fs_info *fs_info, | |
1239 | u64 objectid) | |
1240 | { | |
1241 | struct extent_buffer *leaf; | |
1242 | struct btrfs_root *tree_root = fs_info->tree_root; | |
1243 | struct btrfs_root *root; | |
1244 | struct btrfs_key key; | |
1245 | int ret = 0; | |
6463fe58 | 1246 | uuid_le uuid; |
20897f5c | 1247 | |
74e4d827 | 1248 | root = btrfs_alloc_root(fs_info, GFP_KERNEL); |
20897f5c AJ |
1249 | if (!root) |
1250 | return ERR_PTR(-ENOMEM); | |
1251 | ||
da17066c | 1252 | __setup_root(root, fs_info, objectid); |
20897f5c AJ |
1253 | root->root_key.objectid = objectid; |
1254 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1255 | root->root_key.offset = 0; | |
1256 | ||
4d75f8a9 | 1257 | leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); |
20897f5c AJ |
1258 | if (IS_ERR(leaf)) { |
1259 | ret = PTR_ERR(leaf); | |
1dd05682 | 1260 | leaf = NULL; |
20897f5c AJ |
1261 | goto fail; |
1262 | } | |
1263 | ||
b159fa28 | 1264 | memzero_extent_buffer(leaf, 0, sizeof(struct btrfs_header)); |
20897f5c AJ |
1265 | btrfs_set_header_bytenr(leaf, leaf->start); |
1266 | btrfs_set_header_generation(leaf, trans->transid); | |
1267 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1268 | btrfs_set_header_owner(leaf, objectid); | |
1269 | root->node = leaf; | |
1270 | ||
d24ee97b DS |
1271 | write_extent_buffer_fsid(leaf, fs_info->fsid); |
1272 | write_extent_buffer_chunk_tree_uuid(leaf, fs_info->chunk_tree_uuid); | |
20897f5c AJ |
1273 | btrfs_mark_buffer_dirty(leaf); |
1274 | ||
1275 | root->commit_root = btrfs_root_node(root); | |
27cdeb70 | 1276 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
20897f5c AJ |
1277 | |
1278 | root->root_item.flags = 0; | |
1279 | root->root_item.byte_limit = 0; | |
1280 | btrfs_set_root_bytenr(&root->root_item, leaf->start); | |
1281 | btrfs_set_root_generation(&root->root_item, trans->transid); | |
1282 | btrfs_set_root_level(&root->root_item, 0); | |
1283 | btrfs_set_root_refs(&root->root_item, 1); | |
1284 | btrfs_set_root_used(&root->root_item, leaf->len); | |
1285 | btrfs_set_root_last_snapshot(&root->root_item, 0); | |
1286 | btrfs_set_root_dirid(&root->root_item, 0); | |
6463fe58 SB |
1287 | uuid_le_gen(&uuid); |
1288 | memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE); | |
20897f5c AJ |
1289 | root->root_item.drop_level = 0; |
1290 | ||
1291 | key.objectid = objectid; | |
1292 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1293 | key.offset = 0; | |
1294 | ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); | |
1295 | if (ret) | |
1296 | goto fail; | |
1297 | ||
1298 | btrfs_tree_unlock(leaf); | |
1299 | ||
1dd05682 TI |
1300 | return root; |
1301 | ||
20897f5c | 1302 | fail: |
1dd05682 TI |
1303 | if (leaf) { |
1304 | btrfs_tree_unlock(leaf); | |
59885b39 | 1305 | free_extent_buffer(root->commit_root); |
1dd05682 TI |
1306 | free_extent_buffer(leaf); |
1307 | } | |
1308 | kfree(root); | |
20897f5c | 1309 | |
1dd05682 | 1310 | return ERR_PTR(ret); |
20897f5c AJ |
1311 | } |
1312 | ||
7237f183 YZ |
1313 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
1314 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
1315 | { |
1316 | struct btrfs_root *root; | |
7237f183 | 1317 | struct extent_buffer *leaf; |
e02119d5 | 1318 | |
74e4d827 | 1319 | root = btrfs_alloc_root(fs_info, GFP_NOFS); |
e02119d5 | 1320 | if (!root) |
7237f183 | 1321 | return ERR_PTR(-ENOMEM); |
e02119d5 | 1322 | |
da17066c | 1323 | __setup_root(root, fs_info, BTRFS_TREE_LOG_OBJECTID); |
e02119d5 CM |
1324 | |
1325 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; | |
1326 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1327 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
27cdeb70 | 1328 | |
7237f183 | 1329 | /* |
27cdeb70 MX |
1330 | * DON'T set REF_COWS for log trees |
1331 | * | |
7237f183 YZ |
1332 | * log trees do not get reference counted because they go away |
1333 | * before a real commit is actually done. They do store pointers | |
1334 | * to file data extents, and those reference counts still get | |
1335 | * updated (along with back refs to the log tree). | |
1336 | */ | |
e02119d5 | 1337 | |
4d75f8a9 DS |
1338 | leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, |
1339 | NULL, 0, 0, 0); | |
7237f183 YZ |
1340 | if (IS_ERR(leaf)) { |
1341 | kfree(root); | |
1342 | return ERR_CAST(leaf); | |
1343 | } | |
e02119d5 | 1344 | |
b159fa28 | 1345 | memzero_extent_buffer(leaf, 0, sizeof(struct btrfs_header)); |
5d4f98a2 YZ |
1346 | btrfs_set_header_bytenr(leaf, leaf->start); |
1347 | btrfs_set_header_generation(leaf, trans->transid); | |
1348 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1349 | btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID); | |
7237f183 | 1350 | root->node = leaf; |
e02119d5 | 1351 | |
0b246afa | 1352 | write_extent_buffer_fsid(root->node, fs_info->fsid); |
e02119d5 CM |
1353 | btrfs_mark_buffer_dirty(root->node); |
1354 | btrfs_tree_unlock(root->node); | |
7237f183 YZ |
1355 | return root; |
1356 | } | |
1357 | ||
1358 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
1359 | struct btrfs_fs_info *fs_info) | |
1360 | { | |
1361 | struct btrfs_root *log_root; | |
1362 | ||
1363 | log_root = alloc_log_tree(trans, fs_info); | |
1364 | if (IS_ERR(log_root)) | |
1365 | return PTR_ERR(log_root); | |
1366 | WARN_ON(fs_info->log_root_tree); | |
1367 | fs_info->log_root_tree = log_root; | |
1368 | return 0; | |
1369 | } | |
1370 | ||
1371 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
1372 | struct btrfs_root *root) | |
1373 | { | |
0b246afa | 1374 | struct btrfs_fs_info *fs_info = root->fs_info; |
7237f183 YZ |
1375 | struct btrfs_root *log_root; |
1376 | struct btrfs_inode_item *inode_item; | |
1377 | ||
0b246afa | 1378 | log_root = alloc_log_tree(trans, fs_info); |
7237f183 YZ |
1379 | if (IS_ERR(log_root)) |
1380 | return PTR_ERR(log_root); | |
1381 | ||
1382 | log_root->last_trans = trans->transid; | |
1383 | log_root->root_key.offset = root->root_key.objectid; | |
1384 | ||
1385 | inode_item = &log_root->root_item.inode; | |
3cae210f QW |
1386 | btrfs_set_stack_inode_generation(inode_item, 1); |
1387 | btrfs_set_stack_inode_size(inode_item, 3); | |
1388 | btrfs_set_stack_inode_nlink(inode_item, 1); | |
da17066c | 1389 | btrfs_set_stack_inode_nbytes(inode_item, |
0b246afa | 1390 | fs_info->nodesize); |
3cae210f | 1391 | btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); |
7237f183 | 1392 | |
5d4f98a2 | 1393 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1394 | |
1395 | WARN_ON(root->log_root); | |
1396 | root->log_root = log_root; | |
1397 | root->log_transid = 0; | |
d1433deb | 1398 | root->log_transid_committed = -1; |
257c62e1 | 1399 | root->last_log_commit = 0; |
e02119d5 CM |
1400 | return 0; |
1401 | } | |
1402 | ||
35a3621b SB |
1403 | static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, |
1404 | struct btrfs_key *key) | |
e02119d5 CM |
1405 | { |
1406 | struct btrfs_root *root; | |
1407 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
0f7d52f4 | 1408 | struct btrfs_path *path; |
84234f3a | 1409 | u64 generation; |
cb517eab | 1410 | int ret; |
0f7d52f4 | 1411 | |
cb517eab MX |
1412 | path = btrfs_alloc_path(); |
1413 | if (!path) | |
0f7d52f4 | 1414 | return ERR_PTR(-ENOMEM); |
cb517eab | 1415 | |
74e4d827 | 1416 | root = btrfs_alloc_root(fs_info, GFP_NOFS); |
cb517eab MX |
1417 | if (!root) { |
1418 | ret = -ENOMEM; | |
1419 | goto alloc_fail; | |
0f7d52f4 CM |
1420 | } |
1421 | ||
da17066c | 1422 | __setup_root(root, fs_info, key->objectid); |
0f7d52f4 | 1423 | |
cb517eab MX |
1424 | ret = btrfs_find_root(tree_root, key, path, |
1425 | &root->root_item, &root->root_key); | |
0f7d52f4 | 1426 | if (ret) { |
13a8a7c8 YZ |
1427 | if (ret > 0) |
1428 | ret = -ENOENT; | |
cb517eab | 1429 | goto find_fail; |
0f7d52f4 | 1430 | } |
13a8a7c8 | 1431 | |
84234f3a | 1432 | generation = btrfs_root_generation(&root->root_item); |
2ff7e61e JM |
1433 | root->node = read_tree_block(fs_info, |
1434 | btrfs_root_bytenr(&root->root_item), | |
ce86cd59 | 1435 | generation); |
64c043de LB |
1436 | if (IS_ERR(root->node)) { |
1437 | ret = PTR_ERR(root->node); | |
cb517eab MX |
1438 | goto find_fail; |
1439 | } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) { | |
1440 | ret = -EIO; | |
64c043de LB |
1441 | free_extent_buffer(root->node); |
1442 | goto find_fail; | |
416bc658 | 1443 | } |
5d4f98a2 | 1444 | root->commit_root = btrfs_root_node(root); |
13a8a7c8 | 1445 | out: |
cb517eab MX |
1446 | btrfs_free_path(path); |
1447 | return root; | |
1448 | ||
cb517eab MX |
1449 | find_fail: |
1450 | kfree(root); | |
1451 | alloc_fail: | |
1452 | root = ERR_PTR(ret); | |
1453 | goto out; | |
1454 | } | |
1455 | ||
1456 | struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root, | |
1457 | struct btrfs_key *location) | |
1458 | { | |
1459 | struct btrfs_root *root; | |
1460 | ||
1461 | root = btrfs_read_tree_root(tree_root, location); | |
1462 | if (IS_ERR(root)) | |
1463 | return root; | |
1464 | ||
1465 | if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) { | |
27cdeb70 | 1466 | set_bit(BTRFS_ROOT_REF_COWS, &root->state); |
08fe4db1 LZ |
1467 | btrfs_check_and_init_root_item(&root->root_item); |
1468 | } | |
13a8a7c8 | 1469 | |
5eda7b5e CM |
1470 | return root; |
1471 | } | |
1472 | ||
cb517eab MX |
1473 | int btrfs_init_fs_root(struct btrfs_root *root) |
1474 | { | |
1475 | int ret; | |
8257b2dc | 1476 | struct btrfs_subvolume_writers *writers; |
cb517eab MX |
1477 | |
1478 | root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS); | |
1479 | root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned), | |
1480 | GFP_NOFS); | |
1481 | if (!root->free_ino_pinned || !root->free_ino_ctl) { | |
1482 | ret = -ENOMEM; | |
1483 | goto fail; | |
1484 | } | |
1485 | ||
8257b2dc MX |
1486 | writers = btrfs_alloc_subvolume_writers(); |
1487 | if (IS_ERR(writers)) { | |
1488 | ret = PTR_ERR(writers); | |
1489 | goto fail; | |
1490 | } | |
1491 | root->subv_writers = writers; | |
1492 | ||
cb517eab | 1493 | btrfs_init_free_ino_ctl(root); |
57cdc8db DS |
1494 | spin_lock_init(&root->ino_cache_lock); |
1495 | init_waitqueue_head(&root->ino_cache_wait); | |
cb517eab MX |
1496 | |
1497 | ret = get_anon_bdev(&root->anon_dev); | |
1498 | if (ret) | |
876d2cf1 | 1499 | goto fail; |
f32e48e9 CR |
1500 | |
1501 | mutex_lock(&root->objectid_mutex); | |
1502 | ret = btrfs_find_highest_objectid(root, | |
1503 | &root->highest_objectid); | |
1504 | if (ret) { | |
1505 | mutex_unlock(&root->objectid_mutex); | |
876d2cf1 | 1506 | goto fail; |
f32e48e9 CR |
1507 | } |
1508 | ||
1509 | ASSERT(root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); | |
1510 | ||
1511 | mutex_unlock(&root->objectid_mutex); | |
1512 | ||
cb517eab MX |
1513 | return 0; |
1514 | fail: | |
876d2cf1 | 1515 | /* the caller is responsible to call free_fs_root */ |
cb517eab MX |
1516 | return ret; |
1517 | } | |
1518 | ||
35bbb97f JM |
1519 | struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
1520 | u64 root_id) | |
cb517eab MX |
1521 | { |
1522 | struct btrfs_root *root; | |
1523 | ||
1524 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1525 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
1526 | (unsigned long)root_id); | |
1527 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
1528 | return root; | |
1529 | } | |
1530 | ||
1531 | int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, | |
1532 | struct btrfs_root *root) | |
1533 | { | |
1534 | int ret; | |
1535 | ||
e1860a77 | 1536 | ret = radix_tree_preload(GFP_NOFS); |
cb517eab MX |
1537 | if (ret) |
1538 | return ret; | |
1539 | ||
1540 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1541 | ret = radix_tree_insert(&fs_info->fs_roots_radix, | |
1542 | (unsigned long)root->root_key.objectid, | |
1543 | root); | |
1544 | if (ret == 0) | |
27cdeb70 | 1545 | set_bit(BTRFS_ROOT_IN_RADIX, &root->state); |
cb517eab MX |
1546 | spin_unlock(&fs_info->fs_roots_radix_lock); |
1547 | radix_tree_preload_end(); | |
1548 | ||
1549 | return ret; | |
1550 | } | |
1551 | ||
c00869f1 MX |
1552 | struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, |
1553 | struct btrfs_key *location, | |
1554 | bool check_ref) | |
5eda7b5e CM |
1555 | { |
1556 | struct btrfs_root *root; | |
381cf658 | 1557 | struct btrfs_path *path; |
1d4c08e0 | 1558 | struct btrfs_key key; |
5eda7b5e CM |
1559 | int ret; |
1560 | ||
edbd8d4e CM |
1561 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
1562 | return fs_info->tree_root; | |
1563 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1564 | return fs_info->extent_root; | |
8f18cf13 CM |
1565 | if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) |
1566 | return fs_info->chunk_root; | |
1567 | if (location->objectid == BTRFS_DEV_TREE_OBJECTID) | |
1568 | return fs_info->dev_root; | |
0403e47e YZ |
1569 | if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) |
1570 | return fs_info->csum_root; | |
bcef60f2 AJ |
1571 | if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID) |
1572 | return fs_info->quota_root ? fs_info->quota_root : | |
1573 | ERR_PTR(-ENOENT); | |
f7a81ea4 SB |
1574 | if (location->objectid == BTRFS_UUID_TREE_OBJECTID) |
1575 | return fs_info->uuid_root ? fs_info->uuid_root : | |
1576 | ERR_PTR(-ENOENT); | |
70f6d82e OS |
1577 | if (location->objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) |
1578 | return fs_info->free_space_root ? fs_info->free_space_root : | |
1579 | ERR_PTR(-ENOENT); | |
4df27c4d | 1580 | again: |
cb517eab | 1581 | root = btrfs_lookup_fs_root(fs_info, location->objectid); |
48475471 | 1582 | if (root) { |
c00869f1 | 1583 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) |
48475471 | 1584 | return ERR_PTR(-ENOENT); |
5eda7b5e | 1585 | return root; |
48475471 | 1586 | } |
5eda7b5e | 1587 | |
cb517eab | 1588 | root = btrfs_read_fs_root(fs_info->tree_root, location); |
5eda7b5e CM |
1589 | if (IS_ERR(root)) |
1590 | return root; | |
3394e160 | 1591 | |
c00869f1 | 1592 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
cb517eab | 1593 | ret = -ENOENT; |
581bb050 | 1594 | goto fail; |
35a30d7c | 1595 | } |
581bb050 | 1596 | |
cb517eab | 1597 | ret = btrfs_init_fs_root(root); |
ac08aedf CM |
1598 | if (ret) |
1599 | goto fail; | |
3394e160 | 1600 | |
381cf658 DS |
1601 | path = btrfs_alloc_path(); |
1602 | if (!path) { | |
1603 | ret = -ENOMEM; | |
1604 | goto fail; | |
1605 | } | |
1d4c08e0 DS |
1606 | key.objectid = BTRFS_ORPHAN_OBJECTID; |
1607 | key.type = BTRFS_ORPHAN_ITEM_KEY; | |
1608 | key.offset = location->objectid; | |
1609 | ||
1610 | ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); | |
381cf658 | 1611 | btrfs_free_path(path); |
d68fc57b YZ |
1612 | if (ret < 0) |
1613 | goto fail; | |
1614 | if (ret == 0) | |
27cdeb70 | 1615 | set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); |
d68fc57b | 1616 | |
cb517eab | 1617 | ret = btrfs_insert_fs_root(fs_info, root); |
0f7d52f4 | 1618 | if (ret) { |
4df27c4d YZ |
1619 | if (ret == -EEXIST) { |
1620 | free_fs_root(root); | |
1621 | goto again; | |
1622 | } | |
1623 | goto fail; | |
0f7d52f4 | 1624 | } |
edbd8d4e | 1625 | return root; |
4df27c4d YZ |
1626 | fail: |
1627 | free_fs_root(root); | |
1628 | return ERR_PTR(ret); | |
edbd8d4e CM |
1629 | } |
1630 | ||
04160088 CM |
1631 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) |
1632 | { | |
1633 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
1634 | int ret = 0; | |
04160088 CM |
1635 | struct btrfs_device *device; |
1636 | struct backing_dev_info *bdi; | |
b7967db7 | 1637 | |
1f78160c XG |
1638 | rcu_read_lock(); |
1639 | list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) { | |
dfe25020 CM |
1640 | if (!device->bdev) |
1641 | continue; | |
efa7c9f9 | 1642 | bdi = device->bdev->bd_bdi; |
ff9ea323 | 1643 | if (bdi_congested(bdi, bdi_bits)) { |
04160088 CM |
1644 | ret = 1; |
1645 | break; | |
1646 | } | |
1647 | } | |
1f78160c | 1648 | rcu_read_unlock(); |
04160088 CM |
1649 | return ret; |
1650 | } | |
1651 | ||
8b712842 CM |
1652 | /* |
1653 | * called by the kthread helper functions to finally call the bio end_io | |
1654 | * functions. This is where read checksum verification actually happens | |
1655 | */ | |
1656 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1657 | { |
ce9adaa5 | 1658 | struct bio *bio; |
97eb6b69 | 1659 | struct btrfs_end_io_wq *end_io_wq; |
ce9adaa5 | 1660 | |
97eb6b69 | 1661 | end_io_wq = container_of(work, struct btrfs_end_io_wq, work); |
8b712842 | 1662 | bio = end_io_wq->bio; |
ce9adaa5 | 1663 | |
4e4cbee9 | 1664 | bio->bi_status = end_io_wq->status; |
8b712842 CM |
1665 | bio->bi_private = end_io_wq->private; |
1666 | bio->bi_end_io = end_io_wq->end_io; | |
97eb6b69 | 1667 | kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq); |
4246a0b6 | 1668 | bio_endio(bio); |
44b8bd7e CM |
1669 | } |
1670 | ||
a74a4b97 CM |
1671 | static int cleaner_kthread(void *arg) |
1672 | { | |
1673 | struct btrfs_root *root = arg; | |
0b246afa | 1674 | struct btrfs_fs_info *fs_info = root->fs_info; |
d0278245 | 1675 | int again; |
da288d28 | 1676 | struct btrfs_trans_handle *trans; |
a74a4b97 CM |
1677 | |
1678 | do { | |
d0278245 | 1679 | again = 0; |
a74a4b97 | 1680 | |
d0278245 | 1681 | /* Make the cleaner go to sleep early. */ |
2ff7e61e | 1682 | if (btrfs_need_cleaner_sleep(fs_info)) |
d0278245 MX |
1683 | goto sleep; |
1684 | ||
90c711ab ZB |
1685 | /* |
1686 | * Do not do anything if we might cause open_ctree() to block | |
1687 | * before we have finished mounting the filesystem. | |
1688 | */ | |
0b246afa | 1689 | if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) |
90c711ab ZB |
1690 | goto sleep; |
1691 | ||
0b246afa | 1692 | if (!mutex_trylock(&fs_info->cleaner_mutex)) |
d0278245 MX |
1693 | goto sleep; |
1694 | ||
dc7f370c MX |
1695 | /* |
1696 | * Avoid the problem that we change the status of the fs | |
1697 | * during the above check and trylock. | |
1698 | */ | |
2ff7e61e | 1699 | if (btrfs_need_cleaner_sleep(fs_info)) { |
0b246afa | 1700 | mutex_unlock(&fs_info->cleaner_mutex); |
dc7f370c | 1701 | goto sleep; |
76dda93c | 1702 | } |
a74a4b97 | 1703 | |
0b246afa | 1704 | mutex_lock(&fs_info->cleaner_delayed_iput_mutex); |
2ff7e61e | 1705 | btrfs_run_delayed_iputs(fs_info); |
0b246afa | 1706 | mutex_unlock(&fs_info->cleaner_delayed_iput_mutex); |
c2d6cb16 | 1707 | |
d0278245 | 1708 | again = btrfs_clean_one_deleted_snapshot(root); |
0b246afa | 1709 | mutex_unlock(&fs_info->cleaner_mutex); |
d0278245 MX |
1710 | |
1711 | /* | |
05323cd1 MX |
1712 | * The defragger has dealt with the R/O remount and umount, |
1713 | * needn't do anything special here. | |
d0278245 | 1714 | */ |
0b246afa | 1715 | btrfs_run_defrag_inodes(fs_info); |
67c5e7d4 FM |
1716 | |
1717 | /* | |
1718 | * Acquires fs_info->delete_unused_bgs_mutex to avoid racing | |
1719 | * with relocation (btrfs_relocate_chunk) and relocation | |
1720 | * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) | |
1721 | * after acquiring fs_info->delete_unused_bgs_mutex. So we | |
1722 | * can't hold, nor need to, fs_info->cleaner_mutex when deleting | |
1723 | * unused block groups. | |
1724 | */ | |
0b246afa | 1725 | btrfs_delete_unused_bgs(fs_info); |
d0278245 | 1726 | sleep: |
838fe188 | 1727 | if (!again) { |
a74a4b97 | 1728 | set_current_state(TASK_INTERRUPTIBLE); |
8929ecfa YZ |
1729 | if (!kthread_should_stop()) |
1730 | schedule(); | |
a74a4b97 CM |
1731 | __set_current_state(TASK_RUNNING); |
1732 | } | |
1733 | } while (!kthread_should_stop()); | |
da288d28 FM |
1734 | |
1735 | /* | |
1736 | * Transaction kthread is stopped before us and wakes us up. | |
1737 | * However we might have started a new transaction and COWed some | |
1738 | * tree blocks when deleting unused block groups for example. So | |
1739 | * make sure we commit the transaction we started to have a clean | |
1740 | * shutdown when evicting the btree inode - if it has dirty pages | |
1741 | * when we do the final iput() on it, eviction will trigger a | |
1742 | * writeback for it which will fail with null pointer dereferences | |
1743 | * since work queues and other resources were already released and | |
1744 | * destroyed by the time the iput/eviction/writeback is made. | |
1745 | */ | |
1746 | trans = btrfs_attach_transaction(root); | |
1747 | if (IS_ERR(trans)) { | |
1748 | if (PTR_ERR(trans) != -ENOENT) | |
0b246afa | 1749 | btrfs_err(fs_info, |
da288d28 FM |
1750 | "cleaner transaction attach returned %ld", |
1751 | PTR_ERR(trans)); | |
1752 | } else { | |
1753 | int ret; | |
1754 | ||
3a45bb20 | 1755 | ret = btrfs_commit_transaction(trans); |
da288d28 | 1756 | if (ret) |
0b246afa | 1757 | btrfs_err(fs_info, |
da288d28 FM |
1758 | "cleaner open transaction commit returned %d", |
1759 | ret); | |
1760 | } | |
1761 | ||
a74a4b97 CM |
1762 | return 0; |
1763 | } | |
1764 | ||
1765 | static int transaction_kthread(void *arg) | |
1766 | { | |
1767 | struct btrfs_root *root = arg; | |
0b246afa | 1768 | struct btrfs_fs_info *fs_info = root->fs_info; |
a74a4b97 CM |
1769 | struct btrfs_trans_handle *trans; |
1770 | struct btrfs_transaction *cur; | |
8929ecfa | 1771 | u64 transid; |
a74a4b97 CM |
1772 | unsigned long now; |
1773 | unsigned long delay; | |
914b2007 | 1774 | bool cannot_commit; |
a74a4b97 CM |
1775 | |
1776 | do { | |
914b2007 | 1777 | cannot_commit = false; |
0b246afa JM |
1778 | delay = HZ * fs_info->commit_interval; |
1779 | mutex_lock(&fs_info->transaction_kthread_mutex); | |
a74a4b97 | 1780 | |
0b246afa JM |
1781 | spin_lock(&fs_info->trans_lock); |
1782 | cur = fs_info->running_transaction; | |
a74a4b97 | 1783 | if (!cur) { |
0b246afa | 1784 | spin_unlock(&fs_info->trans_lock); |
a74a4b97 CM |
1785 | goto sleep; |
1786 | } | |
31153d81 | 1787 | |
a74a4b97 | 1788 | now = get_seconds(); |
4a9d8bde | 1789 | if (cur->state < TRANS_STATE_BLOCKED && |
8b87dc17 | 1790 | (now < cur->start_time || |
0b246afa JM |
1791 | now - cur->start_time < fs_info->commit_interval)) { |
1792 | spin_unlock(&fs_info->trans_lock); | |
a74a4b97 CM |
1793 | delay = HZ * 5; |
1794 | goto sleep; | |
1795 | } | |
8929ecfa | 1796 | transid = cur->transid; |
0b246afa | 1797 | spin_unlock(&fs_info->trans_lock); |
56bec294 | 1798 | |
79787eaa | 1799 | /* If the file system is aborted, this will always fail. */ |
354aa0fb | 1800 | trans = btrfs_attach_transaction(root); |
914b2007 | 1801 | if (IS_ERR(trans)) { |
354aa0fb MX |
1802 | if (PTR_ERR(trans) != -ENOENT) |
1803 | cannot_commit = true; | |
79787eaa | 1804 | goto sleep; |
914b2007 | 1805 | } |
8929ecfa | 1806 | if (transid == trans->transid) { |
3a45bb20 | 1807 | btrfs_commit_transaction(trans); |
8929ecfa | 1808 | } else { |
3a45bb20 | 1809 | btrfs_end_transaction(trans); |
8929ecfa | 1810 | } |
a74a4b97 | 1811 | sleep: |
0b246afa JM |
1812 | wake_up_process(fs_info->cleaner_kthread); |
1813 | mutex_unlock(&fs_info->transaction_kthread_mutex); | |
a74a4b97 | 1814 | |
4e121c06 | 1815 | if (unlikely(test_bit(BTRFS_FS_STATE_ERROR, |
0b246afa | 1816 | &fs_info->fs_state))) |
2ff7e61e | 1817 | btrfs_cleanup_transaction(fs_info); |
ce63f891 JK |
1818 | set_current_state(TASK_INTERRUPTIBLE); |
1819 | if (!kthread_should_stop() && | |
0b246afa | 1820 | (!btrfs_transaction_blocked(fs_info) || |
ce63f891 JK |
1821 | cannot_commit)) |
1822 | schedule_timeout(delay); | |
1823 | __set_current_state(TASK_RUNNING); | |
a74a4b97 CM |
1824 | } while (!kthread_should_stop()); |
1825 | return 0; | |
1826 | } | |
1827 | ||
af31f5e5 CM |
1828 | /* |
1829 | * this will find the highest generation in the array of | |
1830 | * root backups. The index of the highest array is returned, | |
1831 | * or -1 if we can't find anything. | |
1832 | * | |
1833 | * We check to make sure the array is valid by comparing the | |
1834 | * generation of the latest root in the array with the generation | |
1835 | * in the super block. If they don't match we pitch it. | |
1836 | */ | |
1837 | static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen) | |
1838 | { | |
1839 | u64 cur; | |
1840 | int newest_index = -1; | |
1841 | struct btrfs_root_backup *root_backup; | |
1842 | int i; | |
1843 | ||
1844 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
1845 | root_backup = info->super_copy->super_roots + i; | |
1846 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1847 | if (cur == newest_gen) | |
1848 | newest_index = i; | |
1849 | } | |
1850 | ||
1851 | /* check to see if we actually wrapped around */ | |
1852 | if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) { | |
1853 | root_backup = info->super_copy->super_roots; | |
1854 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1855 | if (cur == newest_gen) | |
1856 | newest_index = 0; | |
1857 | } | |
1858 | return newest_index; | |
1859 | } | |
1860 | ||
1861 | ||
1862 | /* | |
1863 | * find the oldest backup so we know where to store new entries | |
1864 | * in the backup array. This will set the backup_root_index | |
1865 | * field in the fs_info struct | |
1866 | */ | |
1867 | static void find_oldest_super_backup(struct btrfs_fs_info *info, | |
1868 | u64 newest_gen) | |
1869 | { | |
1870 | int newest_index = -1; | |
1871 | ||
1872 | newest_index = find_newest_super_backup(info, newest_gen); | |
1873 | /* if there was garbage in there, just move along */ | |
1874 | if (newest_index == -1) { | |
1875 | info->backup_root_index = 0; | |
1876 | } else { | |
1877 | info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1878 | } | |
1879 | } | |
1880 | ||
1881 | /* | |
1882 | * copy all the root pointers into the super backup array. | |
1883 | * this will bump the backup pointer by one when it is | |
1884 | * done | |
1885 | */ | |
1886 | static void backup_super_roots(struct btrfs_fs_info *info) | |
1887 | { | |
1888 | int next_backup; | |
1889 | struct btrfs_root_backup *root_backup; | |
1890 | int last_backup; | |
1891 | ||
1892 | next_backup = info->backup_root_index; | |
1893 | last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
1894 | BTRFS_NUM_BACKUP_ROOTS; | |
1895 | ||
1896 | /* | |
1897 | * just overwrite the last backup if we're at the same generation | |
1898 | * this happens only at umount | |
1899 | */ | |
1900 | root_backup = info->super_for_commit->super_roots + last_backup; | |
1901 | if (btrfs_backup_tree_root_gen(root_backup) == | |
1902 | btrfs_header_generation(info->tree_root->node)) | |
1903 | next_backup = last_backup; | |
1904 | ||
1905 | root_backup = info->super_for_commit->super_roots + next_backup; | |
1906 | ||
1907 | /* | |
1908 | * make sure all of our padding and empty slots get zero filled | |
1909 | * regardless of which ones we use today | |
1910 | */ | |
1911 | memset(root_backup, 0, sizeof(*root_backup)); | |
1912 | ||
1913 | info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1914 | ||
1915 | btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); | |
1916 | btrfs_set_backup_tree_root_gen(root_backup, | |
1917 | btrfs_header_generation(info->tree_root->node)); | |
1918 | ||
1919 | btrfs_set_backup_tree_root_level(root_backup, | |
1920 | btrfs_header_level(info->tree_root->node)); | |
1921 | ||
1922 | btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); | |
1923 | btrfs_set_backup_chunk_root_gen(root_backup, | |
1924 | btrfs_header_generation(info->chunk_root->node)); | |
1925 | btrfs_set_backup_chunk_root_level(root_backup, | |
1926 | btrfs_header_level(info->chunk_root->node)); | |
1927 | ||
1928 | btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); | |
1929 | btrfs_set_backup_extent_root_gen(root_backup, | |
1930 | btrfs_header_generation(info->extent_root->node)); | |
1931 | btrfs_set_backup_extent_root_level(root_backup, | |
1932 | btrfs_header_level(info->extent_root->node)); | |
1933 | ||
7c7e82a7 CM |
1934 | /* |
1935 | * we might commit during log recovery, which happens before we set | |
1936 | * the fs_root. Make sure it is valid before we fill it in. | |
1937 | */ | |
1938 | if (info->fs_root && info->fs_root->node) { | |
1939 | btrfs_set_backup_fs_root(root_backup, | |
1940 | info->fs_root->node->start); | |
1941 | btrfs_set_backup_fs_root_gen(root_backup, | |
af31f5e5 | 1942 | btrfs_header_generation(info->fs_root->node)); |
7c7e82a7 | 1943 | btrfs_set_backup_fs_root_level(root_backup, |
af31f5e5 | 1944 | btrfs_header_level(info->fs_root->node)); |
7c7e82a7 | 1945 | } |
af31f5e5 CM |
1946 | |
1947 | btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); | |
1948 | btrfs_set_backup_dev_root_gen(root_backup, | |
1949 | btrfs_header_generation(info->dev_root->node)); | |
1950 | btrfs_set_backup_dev_root_level(root_backup, | |
1951 | btrfs_header_level(info->dev_root->node)); | |
1952 | ||
1953 | btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); | |
1954 | btrfs_set_backup_csum_root_gen(root_backup, | |
1955 | btrfs_header_generation(info->csum_root->node)); | |
1956 | btrfs_set_backup_csum_root_level(root_backup, | |
1957 | btrfs_header_level(info->csum_root->node)); | |
1958 | ||
1959 | btrfs_set_backup_total_bytes(root_backup, | |
1960 | btrfs_super_total_bytes(info->super_copy)); | |
1961 | btrfs_set_backup_bytes_used(root_backup, | |
1962 | btrfs_super_bytes_used(info->super_copy)); | |
1963 | btrfs_set_backup_num_devices(root_backup, | |
1964 | btrfs_super_num_devices(info->super_copy)); | |
1965 | ||
1966 | /* | |
1967 | * if we don't copy this out to the super_copy, it won't get remembered | |
1968 | * for the next commit | |
1969 | */ | |
1970 | memcpy(&info->super_copy->super_roots, | |
1971 | &info->super_for_commit->super_roots, | |
1972 | sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); | |
1973 | } | |
1974 | ||
1975 | /* | |
1976 | * this copies info out of the root backup array and back into | |
1977 | * the in-memory super block. It is meant to help iterate through | |
1978 | * the array, so you send it the number of backups you've already | |
1979 | * tried and the last backup index you used. | |
1980 | * | |
1981 | * this returns -1 when it has tried all the backups | |
1982 | */ | |
1983 | static noinline int next_root_backup(struct btrfs_fs_info *info, | |
1984 | struct btrfs_super_block *super, | |
1985 | int *num_backups_tried, int *backup_index) | |
1986 | { | |
1987 | struct btrfs_root_backup *root_backup; | |
1988 | int newest = *backup_index; | |
1989 | ||
1990 | if (*num_backups_tried == 0) { | |
1991 | u64 gen = btrfs_super_generation(super); | |
1992 | ||
1993 | newest = find_newest_super_backup(info, gen); | |
1994 | if (newest == -1) | |
1995 | return -1; | |
1996 | ||
1997 | *backup_index = newest; | |
1998 | *num_backups_tried = 1; | |
1999 | } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) { | |
2000 | /* we've tried all the backups, all done */ | |
2001 | return -1; | |
2002 | } else { | |
2003 | /* jump to the next oldest backup */ | |
2004 | newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
2005 | BTRFS_NUM_BACKUP_ROOTS; | |
2006 | *backup_index = newest; | |
2007 | *num_backups_tried += 1; | |
2008 | } | |
2009 | root_backup = super->super_roots + newest; | |
2010 | ||
2011 | btrfs_set_super_generation(super, | |
2012 | btrfs_backup_tree_root_gen(root_backup)); | |
2013 | btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); | |
2014 | btrfs_set_super_root_level(super, | |
2015 | btrfs_backup_tree_root_level(root_backup)); | |
2016 | btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); | |
2017 | ||
2018 | /* | |
2019 | * fixme: the total bytes and num_devices need to match or we should | |
2020 | * need a fsck | |
2021 | */ | |
2022 | btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); | |
2023 | btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); | |
2024 | return 0; | |
2025 | } | |
2026 | ||
7abadb64 LB |
2027 | /* helper to cleanup workers */ |
2028 | static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) | |
2029 | { | |
dc6e3209 | 2030 | btrfs_destroy_workqueue(fs_info->fixup_workers); |
afe3d242 | 2031 | btrfs_destroy_workqueue(fs_info->delalloc_workers); |
5cdc7ad3 | 2032 | btrfs_destroy_workqueue(fs_info->workers); |
fccb5d86 | 2033 | btrfs_destroy_workqueue(fs_info->endio_workers); |
fccb5d86 | 2034 | btrfs_destroy_workqueue(fs_info->endio_raid56_workers); |
8b110e39 | 2035 | btrfs_destroy_workqueue(fs_info->endio_repair_workers); |
d05a33ac | 2036 | btrfs_destroy_workqueue(fs_info->rmw_workers); |
fccb5d86 QW |
2037 | btrfs_destroy_workqueue(fs_info->endio_write_workers); |
2038 | btrfs_destroy_workqueue(fs_info->endio_freespace_worker); | |
a8c93d4e | 2039 | btrfs_destroy_workqueue(fs_info->submit_workers); |
5b3bc44e | 2040 | btrfs_destroy_workqueue(fs_info->delayed_workers); |
e66f0bb1 | 2041 | btrfs_destroy_workqueue(fs_info->caching_workers); |
736cfa15 | 2042 | btrfs_destroy_workqueue(fs_info->readahead_workers); |
a44903ab | 2043 | btrfs_destroy_workqueue(fs_info->flush_workers); |
fc97fab0 | 2044 | btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); |
a79b7d4b | 2045 | btrfs_destroy_workqueue(fs_info->extent_workers); |
a9b9477d FM |
2046 | /* |
2047 | * Now that all other work queues are destroyed, we can safely destroy | |
2048 | * the queues used for metadata I/O, since tasks from those other work | |
2049 | * queues can do metadata I/O operations. | |
2050 | */ | |
2051 | btrfs_destroy_workqueue(fs_info->endio_meta_workers); | |
2052 | btrfs_destroy_workqueue(fs_info->endio_meta_write_workers); | |
7abadb64 LB |
2053 | } |
2054 | ||
2e9f5954 R |
2055 | static void free_root_extent_buffers(struct btrfs_root *root) |
2056 | { | |
2057 | if (root) { | |
2058 | free_extent_buffer(root->node); | |
2059 | free_extent_buffer(root->commit_root); | |
2060 | root->node = NULL; | |
2061 | root->commit_root = NULL; | |
2062 | } | |
2063 | } | |
2064 | ||
af31f5e5 CM |
2065 | /* helper to cleanup tree roots */ |
2066 | static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root) | |
2067 | { | |
2e9f5954 | 2068 | free_root_extent_buffers(info->tree_root); |
655b09fe | 2069 | |
2e9f5954 R |
2070 | free_root_extent_buffers(info->dev_root); |
2071 | free_root_extent_buffers(info->extent_root); | |
2072 | free_root_extent_buffers(info->csum_root); | |
2073 | free_root_extent_buffers(info->quota_root); | |
2074 | free_root_extent_buffers(info->uuid_root); | |
2075 | if (chunk_root) | |
2076 | free_root_extent_buffers(info->chunk_root); | |
70f6d82e | 2077 | free_root_extent_buffers(info->free_space_root); |
af31f5e5 CM |
2078 | } |
2079 | ||
faa2dbf0 | 2080 | void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) |
171f6537 JB |
2081 | { |
2082 | int ret; | |
2083 | struct btrfs_root *gang[8]; | |
2084 | int i; | |
2085 | ||
2086 | while (!list_empty(&fs_info->dead_roots)) { | |
2087 | gang[0] = list_entry(fs_info->dead_roots.next, | |
2088 | struct btrfs_root, root_list); | |
2089 | list_del(&gang[0]->root_list); | |
2090 | ||
27cdeb70 | 2091 | if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) { |
cb517eab | 2092 | btrfs_drop_and_free_fs_root(fs_info, gang[0]); |
171f6537 JB |
2093 | } else { |
2094 | free_extent_buffer(gang[0]->node); | |
2095 | free_extent_buffer(gang[0]->commit_root); | |
b0feb9d9 | 2096 | btrfs_put_fs_root(gang[0]); |
171f6537 JB |
2097 | } |
2098 | } | |
2099 | ||
2100 | while (1) { | |
2101 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2102 | (void **)gang, 0, | |
2103 | ARRAY_SIZE(gang)); | |
2104 | if (!ret) | |
2105 | break; | |
2106 | for (i = 0; i < ret; i++) | |
cb517eab | 2107 | btrfs_drop_and_free_fs_root(fs_info, gang[i]); |
171f6537 | 2108 | } |
1a4319cc LB |
2109 | |
2110 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { | |
2111 | btrfs_free_log_root_tree(NULL, fs_info); | |
2ff7e61e | 2112 | btrfs_destroy_pinned_extent(fs_info, fs_info->pinned_extents); |
1a4319cc | 2113 | } |
171f6537 | 2114 | } |
af31f5e5 | 2115 | |
638aa7ed ES |
2116 | static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) |
2117 | { | |
2118 | mutex_init(&fs_info->scrub_lock); | |
2119 | atomic_set(&fs_info->scrubs_running, 0); | |
2120 | atomic_set(&fs_info->scrub_pause_req, 0); | |
2121 | atomic_set(&fs_info->scrubs_paused, 0); | |
2122 | atomic_set(&fs_info->scrub_cancel_req, 0); | |
2123 | init_waitqueue_head(&fs_info->scrub_pause_wait); | |
2124 | fs_info->scrub_workers_refcnt = 0; | |
2125 | } | |
2126 | ||
779a65a4 ES |
2127 | static void btrfs_init_balance(struct btrfs_fs_info *fs_info) |
2128 | { | |
2129 | spin_lock_init(&fs_info->balance_lock); | |
2130 | mutex_init(&fs_info->balance_mutex); | |
2131 | atomic_set(&fs_info->balance_running, 0); | |
2132 | atomic_set(&fs_info->balance_pause_req, 0); | |
2133 | atomic_set(&fs_info->balance_cancel_req, 0); | |
2134 | fs_info->balance_ctl = NULL; | |
2135 | init_waitqueue_head(&fs_info->balance_wait_q); | |
2136 | } | |
2137 | ||
6bccf3ab | 2138 | static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info) |
f37938e0 | 2139 | { |
2ff7e61e JM |
2140 | struct inode *inode = fs_info->btree_inode; |
2141 | ||
2142 | inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; | |
2143 | set_nlink(inode, 1); | |
f37938e0 ES |
2144 | /* |
2145 | * we set the i_size on the btree inode to the max possible int. | |
2146 | * the real end of the address space is determined by all of | |
2147 | * the devices in the system | |
2148 | */ | |
2ff7e61e JM |
2149 | inode->i_size = OFFSET_MAX; |
2150 | inode->i_mapping->a_ops = &btree_aops; | |
f37938e0 | 2151 | |
2ff7e61e | 2152 | RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); |
c6100a4b | 2153 | extent_io_tree_init(&BTRFS_I(inode)->io_tree, inode); |
2ff7e61e JM |
2154 | BTRFS_I(inode)->io_tree.track_uptodate = 0; |
2155 | extent_map_tree_init(&BTRFS_I(inode)->extent_tree); | |
f37938e0 | 2156 | |
2ff7e61e | 2157 | BTRFS_I(inode)->io_tree.ops = &btree_extent_io_ops; |
f37938e0 | 2158 | |
2ff7e61e JM |
2159 | BTRFS_I(inode)->root = fs_info->tree_root; |
2160 | memset(&BTRFS_I(inode)->location, 0, sizeof(struct btrfs_key)); | |
2161 | set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); | |
2162 | btrfs_insert_inode_hash(inode); | |
f37938e0 ES |
2163 | } |
2164 | ||
ad618368 ES |
2165 | static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) |
2166 | { | |
2167 | fs_info->dev_replace.lock_owner = 0; | |
2168 | atomic_set(&fs_info->dev_replace.nesting_level, 0); | |
2169 | mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); | |
73beece9 LB |
2170 | rwlock_init(&fs_info->dev_replace.lock); |
2171 | atomic_set(&fs_info->dev_replace.read_locks, 0); | |
2172 | atomic_set(&fs_info->dev_replace.blocking_readers, 0); | |
ad618368 | 2173 | init_waitqueue_head(&fs_info->replace_wait); |
73beece9 | 2174 | init_waitqueue_head(&fs_info->dev_replace.read_lock_wq); |
ad618368 ES |
2175 | } |
2176 | ||
f9e92e40 ES |
2177 | static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) |
2178 | { | |
2179 | spin_lock_init(&fs_info->qgroup_lock); | |
2180 | mutex_init(&fs_info->qgroup_ioctl_lock); | |
2181 | fs_info->qgroup_tree = RB_ROOT; | |
2182 | fs_info->qgroup_op_tree = RB_ROOT; | |
2183 | INIT_LIST_HEAD(&fs_info->dirty_qgroups); | |
2184 | fs_info->qgroup_seq = 1; | |
f9e92e40 | 2185 | fs_info->qgroup_ulist = NULL; |
d2c609b8 | 2186 | fs_info->qgroup_rescan_running = false; |
f9e92e40 ES |
2187 | mutex_init(&fs_info->qgroup_rescan_lock); |
2188 | } | |
2189 | ||
2a458198 ES |
2190 | static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info, |
2191 | struct btrfs_fs_devices *fs_devices) | |
2192 | { | |
2193 | int max_active = fs_info->thread_pool_size; | |
6f011058 | 2194 | unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; |
2a458198 ES |
2195 | |
2196 | fs_info->workers = | |
cb001095 JM |
2197 | btrfs_alloc_workqueue(fs_info, "worker", |
2198 | flags | WQ_HIGHPRI, max_active, 16); | |
2a458198 ES |
2199 | |
2200 | fs_info->delalloc_workers = | |
cb001095 JM |
2201 | btrfs_alloc_workqueue(fs_info, "delalloc", |
2202 | flags, max_active, 2); | |
2a458198 ES |
2203 | |
2204 | fs_info->flush_workers = | |
cb001095 JM |
2205 | btrfs_alloc_workqueue(fs_info, "flush_delalloc", |
2206 | flags, max_active, 0); | |
2a458198 ES |
2207 | |
2208 | fs_info->caching_workers = | |
cb001095 | 2209 | btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); |
2a458198 ES |
2210 | |
2211 | /* | |
2212 | * a higher idle thresh on the submit workers makes it much more | |
2213 | * likely that bios will be send down in a sane order to the | |
2214 | * devices | |
2215 | */ | |
2216 | fs_info->submit_workers = | |
cb001095 | 2217 | btrfs_alloc_workqueue(fs_info, "submit", flags, |
2a458198 ES |
2218 | min_t(u64, fs_devices->num_devices, |
2219 | max_active), 64); | |
2220 | ||
2221 | fs_info->fixup_workers = | |
cb001095 | 2222 | btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0); |
2a458198 ES |
2223 | |
2224 | /* | |
2225 | * endios are largely parallel and should have a very | |
2226 | * low idle thresh | |
2227 | */ | |
2228 | fs_info->endio_workers = | |
cb001095 | 2229 | btrfs_alloc_workqueue(fs_info, "endio", flags, max_active, 4); |
2a458198 | 2230 | fs_info->endio_meta_workers = |
cb001095 JM |
2231 | btrfs_alloc_workqueue(fs_info, "endio-meta", flags, |
2232 | max_active, 4); | |
2a458198 | 2233 | fs_info->endio_meta_write_workers = |
cb001095 JM |
2234 | btrfs_alloc_workqueue(fs_info, "endio-meta-write", flags, |
2235 | max_active, 2); | |
2a458198 | 2236 | fs_info->endio_raid56_workers = |
cb001095 JM |
2237 | btrfs_alloc_workqueue(fs_info, "endio-raid56", flags, |
2238 | max_active, 4); | |
2a458198 | 2239 | fs_info->endio_repair_workers = |
cb001095 | 2240 | btrfs_alloc_workqueue(fs_info, "endio-repair", flags, 1, 0); |
2a458198 | 2241 | fs_info->rmw_workers = |
cb001095 | 2242 | btrfs_alloc_workqueue(fs_info, "rmw", flags, max_active, 2); |
2a458198 | 2243 | fs_info->endio_write_workers = |
cb001095 JM |
2244 | btrfs_alloc_workqueue(fs_info, "endio-write", flags, |
2245 | max_active, 2); | |
2a458198 | 2246 | fs_info->endio_freespace_worker = |
cb001095 JM |
2247 | btrfs_alloc_workqueue(fs_info, "freespace-write", flags, |
2248 | max_active, 0); | |
2a458198 | 2249 | fs_info->delayed_workers = |
cb001095 JM |
2250 | btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, |
2251 | max_active, 0); | |
2a458198 | 2252 | fs_info->readahead_workers = |
cb001095 JM |
2253 | btrfs_alloc_workqueue(fs_info, "readahead", flags, |
2254 | max_active, 2); | |
2a458198 | 2255 | fs_info->qgroup_rescan_workers = |
cb001095 | 2256 | btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0); |
2a458198 | 2257 | fs_info->extent_workers = |
cb001095 | 2258 | btrfs_alloc_workqueue(fs_info, "extent-refs", flags, |
2a458198 ES |
2259 | min_t(u64, fs_devices->num_devices, |
2260 | max_active), 8); | |
2261 | ||
2262 | if (!(fs_info->workers && fs_info->delalloc_workers && | |
2263 | fs_info->submit_workers && fs_info->flush_workers && | |
2264 | fs_info->endio_workers && fs_info->endio_meta_workers && | |
2265 | fs_info->endio_meta_write_workers && | |
2266 | fs_info->endio_repair_workers && | |
2267 | fs_info->endio_write_workers && fs_info->endio_raid56_workers && | |
2268 | fs_info->endio_freespace_worker && fs_info->rmw_workers && | |
2269 | fs_info->caching_workers && fs_info->readahead_workers && | |
2270 | fs_info->fixup_workers && fs_info->delayed_workers && | |
2271 | fs_info->extent_workers && | |
2272 | fs_info->qgroup_rescan_workers)) { | |
2273 | return -ENOMEM; | |
2274 | } | |
2275 | ||
2276 | return 0; | |
2277 | } | |
2278 | ||
63443bf5 ES |
2279 | static int btrfs_replay_log(struct btrfs_fs_info *fs_info, |
2280 | struct btrfs_fs_devices *fs_devices) | |
2281 | { | |
2282 | int ret; | |
63443bf5 ES |
2283 | struct btrfs_root *log_tree_root; |
2284 | struct btrfs_super_block *disk_super = fs_info->super_copy; | |
2285 | u64 bytenr = btrfs_super_log_root(disk_super); | |
2286 | ||
2287 | if (fs_devices->rw_devices == 0) { | |
f14d104d | 2288 | btrfs_warn(fs_info, "log replay required on RO media"); |
63443bf5 ES |
2289 | return -EIO; |
2290 | } | |
2291 | ||
74e4d827 | 2292 | log_tree_root = btrfs_alloc_root(fs_info, GFP_KERNEL); |
63443bf5 ES |
2293 | if (!log_tree_root) |
2294 | return -ENOMEM; | |
2295 | ||
da17066c | 2296 | __setup_root(log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); |
63443bf5 | 2297 | |
2ff7e61e JM |
2298 | log_tree_root->node = read_tree_block(fs_info, bytenr, |
2299 | fs_info->generation + 1); | |
64c043de | 2300 | if (IS_ERR(log_tree_root->node)) { |
f14d104d | 2301 | btrfs_warn(fs_info, "failed to read log tree"); |
0eeff236 | 2302 | ret = PTR_ERR(log_tree_root->node); |
64c043de | 2303 | kfree(log_tree_root); |
0eeff236 | 2304 | return ret; |
64c043de | 2305 | } else if (!extent_buffer_uptodate(log_tree_root->node)) { |
f14d104d | 2306 | btrfs_err(fs_info, "failed to read log tree"); |
63443bf5 ES |
2307 | free_extent_buffer(log_tree_root->node); |
2308 | kfree(log_tree_root); | |
2309 | return -EIO; | |
2310 | } | |
2311 | /* returns with log_tree_root freed on success */ | |
2312 | ret = btrfs_recover_log_trees(log_tree_root); | |
2313 | if (ret) { | |
0b246afa JM |
2314 | btrfs_handle_fs_error(fs_info, ret, |
2315 | "Failed to recover log tree"); | |
63443bf5 ES |
2316 | free_extent_buffer(log_tree_root->node); |
2317 | kfree(log_tree_root); | |
2318 | return ret; | |
2319 | } | |
2320 | ||
bc98a42c | 2321 | if (sb_rdonly(fs_info->sb)) { |
6bccf3ab | 2322 | ret = btrfs_commit_super(fs_info); |
63443bf5 ES |
2323 | if (ret) |
2324 | return ret; | |
2325 | } | |
2326 | ||
2327 | return 0; | |
2328 | } | |
2329 | ||
6bccf3ab | 2330 | static int btrfs_read_roots(struct btrfs_fs_info *fs_info) |
4bbcaa64 | 2331 | { |
6bccf3ab | 2332 | struct btrfs_root *tree_root = fs_info->tree_root; |
a4f3d2c4 | 2333 | struct btrfs_root *root; |
4bbcaa64 ES |
2334 | struct btrfs_key location; |
2335 | int ret; | |
2336 | ||
6bccf3ab JM |
2337 | BUG_ON(!fs_info->tree_root); |
2338 | ||
4bbcaa64 ES |
2339 | location.objectid = BTRFS_EXTENT_TREE_OBJECTID; |
2340 | location.type = BTRFS_ROOT_ITEM_KEY; | |
2341 | location.offset = 0; | |
2342 | ||
a4f3d2c4 DS |
2343 | root = btrfs_read_tree_root(tree_root, &location); |
2344 | if (IS_ERR(root)) | |
2345 | return PTR_ERR(root); | |
2346 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2347 | fs_info->extent_root = root; | |
4bbcaa64 ES |
2348 | |
2349 | location.objectid = BTRFS_DEV_TREE_OBJECTID; | |
a4f3d2c4 DS |
2350 | root = btrfs_read_tree_root(tree_root, &location); |
2351 | if (IS_ERR(root)) | |
2352 | return PTR_ERR(root); | |
2353 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2354 | fs_info->dev_root = root; | |
4bbcaa64 ES |
2355 | btrfs_init_devices_late(fs_info); |
2356 | ||
2357 | location.objectid = BTRFS_CSUM_TREE_OBJECTID; | |
a4f3d2c4 DS |
2358 | root = btrfs_read_tree_root(tree_root, &location); |
2359 | if (IS_ERR(root)) | |
2360 | return PTR_ERR(root); | |
2361 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2362 | fs_info->csum_root = root; | |
4bbcaa64 ES |
2363 | |
2364 | location.objectid = BTRFS_QUOTA_TREE_OBJECTID; | |
a4f3d2c4 DS |
2365 | root = btrfs_read_tree_root(tree_root, &location); |
2366 | if (!IS_ERR(root)) { | |
2367 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
afcdd129 | 2368 | set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); |
a4f3d2c4 | 2369 | fs_info->quota_root = root; |
4bbcaa64 ES |
2370 | } |
2371 | ||
2372 | location.objectid = BTRFS_UUID_TREE_OBJECTID; | |
a4f3d2c4 DS |
2373 | root = btrfs_read_tree_root(tree_root, &location); |
2374 | if (IS_ERR(root)) { | |
2375 | ret = PTR_ERR(root); | |
4bbcaa64 ES |
2376 | if (ret != -ENOENT) |
2377 | return ret; | |
2378 | } else { | |
a4f3d2c4 DS |
2379 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2380 | fs_info->uuid_root = root; | |
4bbcaa64 ES |
2381 | } |
2382 | ||
70f6d82e OS |
2383 | if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { |
2384 | location.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID; | |
2385 | root = btrfs_read_tree_root(tree_root, &location); | |
2386 | if (IS_ERR(root)) | |
2387 | return PTR_ERR(root); | |
2388 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2389 | fs_info->free_space_root = root; | |
2390 | } | |
2391 | ||
4bbcaa64 ES |
2392 | return 0; |
2393 | } | |
2394 | ||
ad2b2c80 AV |
2395 | int open_ctree(struct super_block *sb, |
2396 | struct btrfs_fs_devices *fs_devices, | |
2397 | char *options) | |
2e635a27 | 2398 | { |
db94535d CM |
2399 | u32 sectorsize; |
2400 | u32 nodesize; | |
87ee04eb | 2401 | u32 stripesize; |
84234f3a | 2402 | u64 generation; |
f2b636e8 | 2403 | u64 features; |
3de4586c | 2404 | struct btrfs_key location; |
a061fc8d | 2405 | struct buffer_head *bh; |
4d34b278 | 2406 | struct btrfs_super_block *disk_super; |
815745cf | 2407 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); |
f84a8bd6 | 2408 | struct btrfs_root *tree_root; |
4d34b278 | 2409 | struct btrfs_root *chunk_root; |
eb60ceac | 2410 | int ret; |
e58ca020 | 2411 | int err = -EINVAL; |
af31f5e5 CM |
2412 | int num_backups_tried = 0; |
2413 | int backup_index = 0; | |
5cdc7ad3 | 2414 | int max_active; |
6675df31 | 2415 | int clear_free_space_tree = 0; |
4543df7e | 2416 | |
74e4d827 DS |
2417 | tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info, GFP_KERNEL); |
2418 | chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info, GFP_KERNEL); | |
cb517eab | 2419 | if (!tree_root || !chunk_root) { |
39279cc3 CM |
2420 | err = -ENOMEM; |
2421 | goto fail; | |
2422 | } | |
76dda93c YZ |
2423 | |
2424 | ret = init_srcu_struct(&fs_info->subvol_srcu); | |
2425 | if (ret) { | |
2426 | err = ret; | |
2427 | goto fail; | |
2428 | } | |
2429 | ||
908c7f19 | 2430 | ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); |
e2d84521 MX |
2431 | if (ret) { |
2432 | err = ret; | |
9e11ceee | 2433 | goto fail_srcu; |
e2d84521 | 2434 | } |
09cbfeaf | 2435 | fs_info->dirty_metadata_batch = PAGE_SIZE * |
e2d84521 MX |
2436 | (1 + ilog2(nr_cpu_ids)); |
2437 | ||
908c7f19 | 2438 | ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); |
963d678b MX |
2439 | if (ret) { |
2440 | err = ret; | |
2441 | goto fail_dirty_metadata_bytes; | |
2442 | } | |
2443 | ||
908c7f19 | 2444 | ret = percpu_counter_init(&fs_info->bio_counter, 0, GFP_KERNEL); |
c404e0dc MX |
2445 | if (ret) { |
2446 | err = ret; | |
2447 | goto fail_delalloc_bytes; | |
2448 | } | |
2449 | ||
76dda93c | 2450 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); |
f28491e0 | 2451 | INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); |
8fd17795 | 2452 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 2453 | INIT_LIST_HEAD(&fs_info->dead_roots); |
24bbcf04 | 2454 | INIT_LIST_HEAD(&fs_info->delayed_iputs); |
eb73c1b7 | 2455 | INIT_LIST_HEAD(&fs_info->delalloc_roots); |
11833d66 | 2456 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
eb73c1b7 | 2457 | spin_lock_init(&fs_info->delalloc_root_lock); |
a4abeea4 | 2458 | spin_lock_init(&fs_info->trans_lock); |
76dda93c | 2459 | spin_lock_init(&fs_info->fs_roots_radix_lock); |
24bbcf04 | 2460 | spin_lock_init(&fs_info->delayed_iput_lock); |
4cb5300b | 2461 | spin_lock_init(&fs_info->defrag_inodes_lock); |
f29021b2 | 2462 | spin_lock_init(&fs_info->tree_mod_seq_lock); |
ceda0864 | 2463 | spin_lock_init(&fs_info->super_lock); |
fcebe456 | 2464 | spin_lock_init(&fs_info->qgroup_op_lock); |
f28491e0 | 2465 | spin_lock_init(&fs_info->buffer_lock); |
47ab2a6c | 2466 | spin_lock_init(&fs_info->unused_bgs_lock); |
f29021b2 | 2467 | rwlock_init(&fs_info->tree_mod_log_lock); |
d7c15171 | 2468 | mutex_init(&fs_info->unused_bg_unpin_mutex); |
67c5e7d4 | 2469 | mutex_init(&fs_info->delete_unused_bgs_mutex); |
7585717f | 2470 | mutex_init(&fs_info->reloc_mutex); |
573bfb72 | 2471 | mutex_init(&fs_info->delalloc_root_mutex); |
c2d6cb16 | 2472 | mutex_init(&fs_info->cleaner_delayed_iput_mutex); |
de98ced9 | 2473 | seqlock_init(&fs_info->profiles_lock); |
19c00ddc | 2474 | |
0b86a832 | 2475 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 2476 | INIT_LIST_HEAD(&fs_info->space_info); |
f29021b2 | 2477 | INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); |
47ab2a6c | 2478 | INIT_LIST_HEAD(&fs_info->unused_bgs); |
0b86a832 | 2479 | btrfs_mapping_init(&fs_info->mapping_tree); |
66d8f3dd MX |
2480 | btrfs_init_block_rsv(&fs_info->global_block_rsv, |
2481 | BTRFS_BLOCK_RSV_GLOBAL); | |
66d8f3dd MX |
2482 | btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); |
2483 | btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); | |
2484 | btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); | |
2485 | btrfs_init_block_rsv(&fs_info->delayed_block_rsv, | |
2486 | BTRFS_BLOCK_RSV_DELOPS); | |
771ed689 | 2487 | atomic_set(&fs_info->async_delalloc_pages, 0); |
4cb5300b | 2488 | atomic_set(&fs_info->defrag_running, 0); |
fcebe456 | 2489 | atomic_set(&fs_info->qgroup_op_seq, 0); |
2fefd558 | 2490 | atomic_set(&fs_info->reada_works_cnt, 0); |
fc36ed7e | 2491 | atomic64_set(&fs_info->tree_mod_seq, 0); |
e20d96d6 | 2492 | fs_info->sb = sb; |
95ac567a | 2493 | fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; |
9ed74f2d | 2494 | fs_info->metadata_ratio = 0; |
4cb5300b | 2495 | fs_info->defrag_inodes = RB_ROOT; |
a5ed45f8 | 2496 | atomic64_set(&fs_info->free_chunk_space, 0); |
f29021b2 | 2497 | fs_info->tree_mod_log = RB_ROOT; |
8b87dc17 | 2498 | fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; |
f8c269d7 | 2499 | fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ |
90519d66 | 2500 | /* readahead state */ |
d0164adc | 2501 | INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_DIRECT_RECLAIM); |
90519d66 | 2502 | spin_lock_init(&fs_info->reada_lock); |
fd708b81 | 2503 | btrfs_init_ref_verify(fs_info); |
c8b97818 | 2504 | |
b34b086c CM |
2505 | fs_info->thread_pool_size = min_t(unsigned long, |
2506 | num_online_cpus() + 2, 8); | |
0afbaf8c | 2507 | |
199c2a9c MX |
2508 | INIT_LIST_HEAD(&fs_info->ordered_roots); |
2509 | spin_lock_init(&fs_info->ordered_root_lock); | |
69fe2d75 JB |
2510 | |
2511 | fs_info->btree_inode = new_inode(sb); | |
2512 | if (!fs_info->btree_inode) { | |
2513 | err = -ENOMEM; | |
2514 | goto fail_bio_counter; | |
2515 | } | |
2516 | mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); | |
2517 | ||
16cdcec7 | 2518 | fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), |
74e4d827 | 2519 | GFP_KERNEL); |
16cdcec7 MX |
2520 | if (!fs_info->delayed_root) { |
2521 | err = -ENOMEM; | |
2522 | goto fail_iput; | |
2523 | } | |
2524 | btrfs_init_delayed_root(fs_info->delayed_root); | |
3eaa2885 | 2525 | |
638aa7ed | 2526 | btrfs_init_scrub(fs_info); |
21adbd5c SB |
2527 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
2528 | fs_info->check_integrity_print_mask = 0; | |
2529 | #endif | |
779a65a4 | 2530 | btrfs_init_balance(fs_info); |
21c7e756 | 2531 | btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work); |
a2de733c | 2532 | |
9f6d2510 DS |
2533 | sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; |
2534 | sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); | |
a061fc8d | 2535 | |
6bccf3ab | 2536 | btrfs_init_btree_inode(fs_info); |
76dda93c | 2537 | |
0f9dd46c | 2538 | spin_lock_init(&fs_info->block_group_cache_lock); |
6bef4d31 | 2539 | fs_info->block_group_cache_tree = RB_ROOT; |
a1897fdd | 2540 | fs_info->first_logical_byte = (u64)-1; |
0f9dd46c | 2541 | |
c6100a4b JB |
2542 | extent_io_tree_init(&fs_info->freed_extents[0], NULL); |
2543 | extent_io_tree_init(&fs_info->freed_extents[1], NULL); | |
11833d66 | 2544 | fs_info->pinned_extents = &fs_info->freed_extents[0]; |
afcdd129 | 2545 | set_bit(BTRFS_FS_BARRIER, &fs_info->flags); |
39279cc3 | 2546 | |
5a3f23d5 | 2547 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 2548 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 2549 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
2550 | mutex_init(&fs_info->transaction_kthread_mutex); |
2551 | mutex_init(&fs_info->cleaner_mutex); | |
7d9eb12c | 2552 | mutex_init(&fs_info->volume_mutex); |
1bbc621e | 2553 | mutex_init(&fs_info->ro_block_group_mutex); |
9e351cc8 | 2554 | init_rwsem(&fs_info->commit_root_sem); |
c71bf099 | 2555 | init_rwsem(&fs_info->cleanup_work_sem); |
76dda93c | 2556 | init_rwsem(&fs_info->subvol_sem); |
803b2f54 | 2557 | sema_init(&fs_info->uuid_tree_rescan_sem, 1); |
fa9c0d79 | 2558 | |
ad618368 | 2559 | btrfs_init_dev_replace_locks(fs_info); |
f9e92e40 | 2560 | btrfs_init_qgroup(fs_info); |
416ac51d | 2561 | |
fa9c0d79 CM |
2562 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); |
2563 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
2564 | ||
e6dcd2dc | 2565 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 2566 | init_waitqueue_head(&fs_info->transaction_wait); |
bb9c12c9 | 2567 | init_waitqueue_head(&fs_info->transaction_blocked_wait); |
4854ddd0 | 2568 | init_waitqueue_head(&fs_info->async_submit_wait); |
3768f368 | 2569 | |
04216820 FM |
2570 | INIT_LIST_HEAD(&fs_info->pinned_chunks); |
2571 | ||
da17066c JM |
2572 | /* Usable values until the real ones are cached from the superblock */ |
2573 | fs_info->nodesize = 4096; | |
2574 | fs_info->sectorsize = 4096; | |
2575 | fs_info->stripesize = 4096; | |
2576 | ||
53b381b3 DW |
2577 | ret = btrfs_alloc_stripe_hash_table(fs_info); |
2578 | if (ret) { | |
83c8266a | 2579 | err = ret; |
53b381b3 DW |
2580 | goto fail_alloc; |
2581 | } | |
2582 | ||
da17066c | 2583 | __setup_root(tree_root, fs_info, BTRFS_ROOT_TREE_OBJECTID); |
7eccb903 | 2584 | |
3c4bb26b | 2585 | invalidate_bdev(fs_devices->latest_bdev); |
1104a885 DS |
2586 | |
2587 | /* | |
2588 | * Read super block and check the signature bytes only | |
2589 | */ | |
a512bbf8 | 2590 | bh = btrfs_read_dev_super(fs_devices->latest_bdev); |
92fc03fb AJ |
2591 | if (IS_ERR(bh)) { |
2592 | err = PTR_ERR(bh); | |
16cdcec7 | 2593 | goto fail_alloc; |
20b45077 | 2594 | } |
39279cc3 | 2595 | |
1104a885 DS |
2596 | /* |
2597 | * We want to check superblock checksum, the type is stored inside. | |
2598 | * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). | |
2599 | */ | |
ab8d0fc4 | 2600 | if (btrfs_check_super_csum(fs_info, bh->b_data)) { |
05135f59 | 2601 | btrfs_err(fs_info, "superblock checksum mismatch"); |
1104a885 | 2602 | err = -EINVAL; |
b2acdddf | 2603 | brelse(bh); |
1104a885 DS |
2604 | goto fail_alloc; |
2605 | } | |
2606 | ||
2607 | /* | |
2608 | * super_copy is zeroed at allocation time and we never touch the | |
2609 | * following bytes up to INFO_SIZE, the checksum is calculated from | |
2610 | * the whole block of INFO_SIZE | |
2611 | */ | |
6c41761f DS |
2612 | memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy)); |
2613 | memcpy(fs_info->super_for_commit, fs_info->super_copy, | |
2614 | sizeof(*fs_info->super_for_commit)); | |
a061fc8d | 2615 | brelse(bh); |
5f39d397 | 2616 | |
6c41761f | 2617 | memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE); |
0b86a832 | 2618 | |
3d3a126a | 2619 | ret = btrfs_check_super_valid(fs_info); |
1104a885 | 2620 | if (ret) { |
05135f59 | 2621 | btrfs_err(fs_info, "superblock contains fatal errors"); |
1104a885 DS |
2622 | err = -EINVAL; |
2623 | goto fail_alloc; | |
2624 | } | |
2625 | ||
6c41761f | 2626 | disk_super = fs_info->super_copy; |
0f7d52f4 | 2627 | if (!btrfs_super_root(disk_super)) |
16cdcec7 | 2628 | goto fail_alloc; |
0f7d52f4 | 2629 | |
acce952b | 2630 | /* check FS state, whether FS is broken. */ |
87533c47 MX |
2631 | if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) |
2632 | set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); | |
acce952b | 2633 | |
af31f5e5 CM |
2634 | /* |
2635 | * run through our array of backup supers and setup | |
2636 | * our ring pointer to the oldest one | |
2637 | */ | |
2638 | generation = btrfs_super_generation(disk_super); | |
2639 | find_oldest_super_backup(fs_info, generation); | |
2640 | ||
75e7cb7f LB |
2641 | /* |
2642 | * In the long term, we'll store the compression type in the super | |
2643 | * block, and it'll be used for per file compression control. | |
2644 | */ | |
2645 | fs_info->compress_type = BTRFS_COMPRESS_ZLIB; | |
2646 | ||
2ff7e61e | 2647 | ret = btrfs_parse_options(fs_info, options, sb->s_flags); |
2b82032c YZ |
2648 | if (ret) { |
2649 | err = ret; | |
16cdcec7 | 2650 | goto fail_alloc; |
2b82032c | 2651 | } |
dfe25020 | 2652 | |
f2b636e8 JB |
2653 | features = btrfs_super_incompat_flags(disk_super) & |
2654 | ~BTRFS_FEATURE_INCOMPAT_SUPP; | |
2655 | if (features) { | |
05135f59 DS |
2656 | btrfs_err(fs_info, |
2657 | "cannot mount because of unsupported optional features (%llx)", | |
2658 | features); | |
f2b636e8 | 2659 | err = -EINVAL; |
16cdcec7 | 2660 | goto fail_alloc; |
f2b636e8 JB |
2661 | } |
2662 | ||
5d4f98a2 | 2663 | features = btrfs_super_incompat_flags(disk_super); |
a6fa6fae | 2664 | features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; |
0b246afa | 2665 | if (fs_info->compress_type == BTRFS_COMPRESS_LZO) |
a6fa6fae | 2666 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; |
5c1aab1d NT |
2667 | else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) |
2668 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; | |
727011e0 | 2669 | |
3173a18f | 2670 | if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) |
05135f59 | 2671 | btrfs_info(fs_info, "has skinny extents"); |
3173a18f | 2672 | |
727011e0 CM |
2673 | /* |
2674 | * flag our filesystem as having big metadata blocks if | |
2675 | * they are bigger than the page size | |
2676 | */ | |
09cbfeaf | 2677 | if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) { |
727011e0 | 2678 | if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) |
05135f59 DS |
2679 | btrfs_info(fs_info, |
2680 | "flagging fs with big metadata feature"); | |
727011e0 CM |
2681 | features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; |
2682 | } | |
2683 | ||
bc3f116f | 2684 | nodesize = btrfs_super_nodesize(disk_super); |
bc3f116f | 2685 | sectorsize = btrfs_super_sectorsize(disk_super); |
b7f67055 | 2686 | stripesize = sectorsize; |
707e8a07 | 2687 | fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); |
963d678b | 2688 | fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); |
bc3f116f | 2689 | |
da17066c JM |
2690 | /* Cache block sizes */ |
2691 | fs_info->nodesize = nodesize; | |
2692 | fs_info->sectorsize = sectorsize; | |
2693 | fs_info->stripesize = stripesize; | |
2694 | ||
bc3f116f CM |
2695 | /* |
2696 | * mixed block groups end up with duplicate but slightly offset | |
2697 | * extent buffers for the same range. It leads to corruptions | |
2698 | */ | |
2699 | if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && | |
707e8a07 | 2700 | (sectorsize != nodesize)) { |
05135f59 DS |
2701 | btrfs_err(fs_info, |
2702 | "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", | |
2703 | nodesize, sectorsize); | |
bc3f116f CM |
2704 | goto fail_alloc; |
2705 | } | |
2706 | ||
ceda0864 MX |
2707 | /* |
2708 | * Needn't use the lock because there is no other task which will | |
2709 | * update the flag. | |
2710 | */ | |
a6fa6fae | 2711 | btrfs_set_super_incompat_flags(disk_super, features); |
5d4f98a2 | 2712 | |
f2b636e8 JB |
2713 | features = btrfs_super_compat_ro_flags(disk_super) & |
2714 | ~BTRFS_FEATURE_COMPAT_RO_SUPP; | |
bc98a42c | 2715 | if (!sb_rdonly(sb) && features) { |
05135f59 DS |
2716 | btrfs_err(fs_info, |
2717 | "cannot mount read-write because of unsupported optional features (%llx)", | |
c1c9ff7c | 2718 | features); |
f2b636e8 | 2719 | err = -EINVAL; |
16cdcec7 | 2720 | goto fail_alloc; |
f2b636e8 | 2721 | } |
61d92c32 | 2722 | |
5cdc7ad3 | 2723 | max_active = fs_info->thread_pool_size; |
61d92c32 | 2724 | |
2a458198 ES |
2725 | ret = btrfs_init_workqueues(fs_info, fs_devices); |
2726 | if (ret) { | |
2727 | err = ret; | |
0dc3b84a JB |
2728 | goto fail_sb_buffer; |
2729 | } | |
4543df7e | 2730 | |
9e11ceee JK |
2731 | sb->s_bdi->congested_fn = btrfs_congested_fn; |
2732 | sb->s_bdi->congested_data = fs_info; | |
2733 | sb->s_bdi->capabilities |= BDI_CAP_CGROUP_WRITEBACK; | |
d4417e22 | 2734 | sb->s_bdi->ra_pages = VM_MAX_READAHEAD * SZ_1K / PAGE_SIZE; |
9e11ceee JK |
2735 | sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); |
2736 | sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); | |
4575c9cc | 2737 | |
a061fc8d CM |
2738 | sb->s_blocksize = sectorsize; |
2739 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
ee87cf5e | 2740 | memcpy(&sb->s_uuid, fs_info->fsid, BTRFS_FSID_SIZE); |
db94535d | 2741 | |
925baedd | 2742 | mutex_lock(&fs_info->chunk_mutex); |
6bccf3ab | 2743 | ret = btrfs_read_sys_array(fs_info); |
925baedd | 2744 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 2745 | if (ret) { |
05135f59 | 2746 | btrfs_err(fs_info, "failed to read the system array: %d", ret); |
5d4f98a2 | 2747 | goto fail_sb_buffer; |
84eed90f | 2748 | } |
0b86a832 | 2749 | |
84234f3a | 2750 | generation = btrfs_super_chunk_root_generation(disk_super); |
0b86a832 | 2751 | |
da17066c | 2752 | __setup_root(chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); |
0b86a832 | 2753 | |
2ff7e61e | 2754 | chunk_root->node = read_tree_block(fs_info, |
0b86a832 | 2755 | btrfs_super_chunk_root(disk_super), |
ce86cd59 | 2756 | generation); |
64c043de LB |
2757 | if (IS_ERR(chunk_root->node) || |
2758 | !extent_buffer_uptodate(chunk_root->node)) { | |
05135f59 | 2759 | btrfs_err(fs_info, "failed to read chunk root"); |
e5fffbac | 2760 | if (!IS_ERR(chunk_root->node)) |
2761 | free_extent_buffer(chunk_root->node); | |
95ab1f64 | 2762 | chunk_root->node = NULL; |
af31f5e5 | 2763 | goto fail_tree_roots; |
83121942 | 2764 | } |
5d4f98a2 YZ |
2765 | btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); |
2766 | chunk_root->commit_root = btrfs_root_node(chunk_root); | |
0b86a832 | 2767 | |
e17cade2 | 2768 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
b308bc2f | 2769 | btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE); |
e17cade2 | 2770 | |
5b4aacef | 2771 | ret = btrfs_read_chunk_tree(fs_info); |
2b82032c | 2772 | if (ret) { |
05135f59 | 2773 | btrfs_err(fs_info, "failed to read chunk tree: %d", ret); |
af31f5e5 | 2774 | goto fail_tree_roots; |
2b82032c | 2775 | } |
0b86a832 | 2776 | |
8dabb742 SB |
2777 | /* |
2778 | * keep the device that is marked to be the target device for the | |
2779 | * dev_replace procedure | |
2780 | */ | |
9eaed21e | 2781 | btrfs_close_extra_devices(fs_devices, 0); |
dfe25020 | 2782 | |
a6b0d5c8 | 2783 | if (!fs_devices->latest_bdev) { |
05135f59 | 2784 | btrfs_err(fs_info, "failed to read devices"); |
a6b0d5c8 CM |
2785 | goto fail_tree_roots; |
2786 | } | |
2787 | ||
af31f5e5 | 2788 | retry_root_backup: |
84234f3a | 2789 | generation = btrfs_super_generation(disk_super); |
0b86a832 | 2790 | |
2ff7e61e | 2791 | tree_root->node = read_tree_block(fs_info, |
db94535d | 2792 | btrfs_super_root(disk_super), |
ce86cd59 | 2793 | generation); |
64c043de LB |
2794 | if (IS_ERR(tree_root->node) || |
2795 | !extent_buffer_uptodate(tree_root->node)) { | |
05135f59 | 2796 | btrfs_warn(fs_info, "failed to read tree root"); |
e5fffbac | 2797 | if (!IS_ERR(tree_root->node)) |
2798 | free_extent_buffer(tree_root->node); | |
95ab1f64 | 2799 | tree_root->node = NULL; |
af31f5e5 | 2800 | goto recovery_tree_root; |
83121942 | 2801 | } |
af31f5e5 | 2802 | |
5d4f98a2 YZ |
2803 | btrfs_set_root_node(&tree_root->root_item, tree_root->node); |
2804 | tree_root->commit_root = btrfs_root_node(tree_root); | |
69e9c6c6 | 2805 | btrfs_set_root_refs(&tree_root->root_item, 1); |
db94535d | 2806 | |
f32e48e9 CR |
2807 | mutex_lock(&tree_root->objectid_mutex); |
2808 | ret = btrfs_find_highest_objectid(tree_root, | |
2809 | &tree_root->highest_objectid); | |
2810 | if (ret) { | |
2811 | mutex_unlock(&tree_root->objectid_mutex); | |
2812 | goto recovery_tree_root; | |
2813 | } | |
2814 | ||
2815 | ASSERT(tree_root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); | |
2816 | ||
2817 | mutex_unlock(&tree_root->objectid_mutex); | |
2818 | ||
6bccf3ab | 2819 | ret = btrfs_read_roots(fs_info); |
4bbcaa64 | 2820 | if (ret) |
af31f5e5 | 2821 | goto recovery_tree_root; |
f7a81ea4 | 2822 | |
8929ecfa YZ |
2823 | fs_info->generation = generation; |
2824 | fs_info->last_trans_committed = generation; | |
8929ecfa | 2825 | |
68310a5e ID |
2826 | ret = btrfs_recover_balance(fs_info); |
2827 | if (ret) { | |
05135f59 | 2828 | btrfs_err(fs_info, "failed to recover balance: %d", ret); |
68310a5e ID |
2829 | goto fail_block_groups; |
2830 | } | |
2831 | ||
733f4fbb SB |
2832 | ret = btrfs_init_dev_stats(fs_info); |
2833 | if (ret) { | |
05135f59 | 2834 | btrfs_err(fs_info, "failed to init dev_stats: %d", ret); |
733f4fbb SB |
2835 | goto fail_block_groups; |
2836 | } | |
2837 | ||
8dabb742 SB |
2838 | ret = btrfs_init_dev_replace(fs_info); |
2839 | if (ret) { | |
05135f59 | 2840 | btrfs_err(fs_info, "failed to init dev_replace: %d", ret); |
8dabb742 SB |
2841 | goto fail_block_groups; |
2842 | } | |
2843 | ||
9eaed21e | 2844 | btrfs_close_extra_devices(fs_devices, 1); |
8dabb742 | 2845 | |
b7c35e81 AJ |
2846 | ret = btrfs_sysfs_add_fsid(fs_devices, NULL); |
2847 | if (ret) { | |
05135f59 DS |
2848 | btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", |
2849 | ret); | |
b7c35e81 AJ |
2850 | goto fail_block_groups; |
2851 | } | |
2852 | ||
2853 | ret = btrfs_sysfs_add_device(fs_devices); | |
2854 | if (ret) { | |
05135f59 DS |
2855 | btrfs_err(fs_info, "failed to init sysfs device interface: %d", |
2856 | ret); | |
b7c35e81 AJ |
2857 | goto fail_fsdev_sysfs; |
2858 | } | |
2859 | ||
96f3136e | 2860 | ret = btrfs_sysfs_add_mounted(fs_info); |
c59021f8 | 2861 | if (ret) { |
05135f59 | 2862 | btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); |
b7c35e81 | 2863 | goto fail_fsdev_sysfs; |
c59021f8 | 2864 | } |
2865 | ||
c59021f8 | 2866 | ret = btrfs_init_space_info(fs_info); |
2867 | if (ret) { | |
05135f59 | 2868 | btrfs_err(fs_info, "failed to initialize space info: %d", ret); |
2365dd3c | 2869 | goto fail_sysfs; |
c59021f8 | 2870 | } |
2871 | ||
5b4aacef | 2872 | ret = btrfs_read_block_groups(fs_info); |
1b1d1f66 | 2873 | if (ret) { |
05135f59 | 2874 | btrfs_err(fs_info, "failed to read block groups: %d", ret); |
2365dd3c | 2875 | goto fail_sysfs; |
1b1d1f66 | 2876 | } |
4330e183 | 2877 | |
0f0d1272 | 2878 | if (!sb_rdonly(sb) && !btrfs_check_rw_degradable(fs_info)) { |
05135f59 | 2879 | btrfs_warn(fs_info, |
4330e183 | 2880 | "writeable mount is not allowed due to too many missing devices"); |
2365dd3c | 2881 | goto fail_sysfs; |
292fd7fc | 2882 | } |
9078a3e1 | 2883 | |
a74a4b97 CM |
2884 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
2885 | "btrfs-cleaner"); | |
57506d50 | 2886 | if (IS_ERR(fs_info->cleaner_kthread)) |
2365dd3c | 2887 | goto fail_sysfs; |
a74a4b97 CM |
2888 | |
2889 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
2890 | tree_root, | |
2891 | "btrfs-transaction"); | |
57506d50 | 2892 | if (IS_ERR(fs_info->transaction_kthread)) |
3f157a2f | 2893 | goto fail_cleaner; |
a74a4b97 | 2894 | |
583b7231 | 2895 | if (!btrfs_test_opt(fs_info, NOSSD) && |
c289811c | 2896 | !fs_info->fs_devices->rotating) { |
583b7231 | 2897 | btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); |
c289811c CM |
2898 | } |
2899 | ||
572d9ab7 | 2900 | /* |
01327610 | 2901 | * Mount does not set all options immediately, we can do it now and do |
572d9ab7 DS |
2902 | * not have to wait for transaction commit |
2903 | */ | |
2904 | btrfs_apply_pending_changes(fs_info); | |
3818aea2 | 2905 | |
21adbd5c | 2906 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
0b246afa | 2907 | if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { |
2ff7e61e | 2908 | ret = btrfsic_mount(fs_info, fs_devices, |
0b246afa | 2909 | btrfs_test_opt(fs_info, |
21adbd5c SB |
2910 | CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? |
2911 | 1 : 0, | |
2912 | fs_info->check_integrity_print_mask); | |
2913 | if (ret) | |
05135f59 DS |
2914 | btrfs_warn(fs_info, |
2915 | "failed to initialize integrity check module: %d", | |
2916 | ret); | |
21adbd5c SB |
2917 | } |
2918 | #endif | |
bcef60f2 AJ |
2919 | ret = btrfs_read_qgroup_config(fs_info); |
2920 | if (ret) | |
2921 | goto fail_trans_kthread; | |
21adbd5c | 2922 | |
fd708b81 JB |
2923 | if (btrfs_build_ref_tree(fs_info)) |
2924 | btrfs_err(fs_info, "couldn't build ref tree"); | |
2925 | ||
96da0919 QW |
2926 | /* do not make disk changes in broken FS or nologreplay is given */ |
2927 | if (btrfs_super_log_root(disk_super) != 0 && | |
0b246afa | 2928 | !btrfs_test_opt(fs_info, NOLOGREPLAY)) { |
63443bf5 | 2929 | ret = btrfs_replay_log(fs_info, fs_devices); |
79787eaa | 2930 | if (ret) { |
63443bf5 | 2931 | err = ret; |
28c16cbb | 2932 | goto fail_qgroup; |
79787eaa | 2933 | } |
e02119d5 | 2934 | } |
1a40e23b | 2935 | |
6bccf3ab | 2936 | ret = btrfs_find_orphan_roots(fs_info); |
79787eaa | 2937 | if (ret) |
28c16cbb | 2938 | goto fail_qgroup; |
76dda93c | 2939 | |
bc98a42c | 2940 | if (!sb_rdonly(sb)) { |
d68fc57b | 2941 | ret = btrfs_cleanup_fs_roots(fs_info); |
44c44af2 | 2942 | if (ret) |
28c16cbb | 2943 | goto fail_qgroup; |
90c711ab ZB |
2944 | |
2945 | mutex_lock(&fs_info->cleaner_mutex); | |
5d4f98a2 | 2946 | ret = btrfs_recover_relocation(tree_root); |
90c711ab | 2947 | mutex_unlock(&fs_info->cleaner_mutex); |
d7ce5843 | 2948 | if (ret < 0) { |
05135f59 DS |
2949 | btrfs_warn(fs_info, "failed to recover relocation: %d", |
2950 | ret); | |
d7ce5843 | 2951 | err = -EINVAL; |
bcef60f2 | 2952 | goto fail_qgroup; |
d7ce5843 | 2953 | } |
7c2ca468 | 2954 | } |
1a40e23b | 2955 | |
3de4586c CM |
2956 | location.objectid = BTRFS_FS_TREE_OBJECTID; |
2957 | location.type = BTRFS_ROOT_ITEM_KEY; | |
cb517eab | 2958 | location.offset = 0; |
3de4586c | 2959 | |
3de4586c | 2960 | fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); |
3140c9a3 DC |
2961 | if (IS_ERR(fs_info->fs_root)) { |
2962 | err = PTR_ERR(fs_info->fs_root); | |
bcef60f2 | 2963 | goto fail_qgroup; |
3140c9a3 | 2964 | } |
c289811c | 2965 | |
bc98a42c | 2966 | if (sb_rdonly(sb)) |
2b6ba629 | 2967 | return 0; |
59641015 | 2968 | |
f8d468a1 OS |
2969 | if (btrfs_test_opt(fs_info, CLEAR_CACHE) && |
2970 | btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
6675df31 OS |
2971 | clear_free_space_tree = 1; |
2972 | } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && | |
2973 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { | |
2974 | btrfs_warn(fs_info, "free space tree is invalid"); | |
2975 | clear_free_space_tree = 1; | |
2976 | } | |
2977 | ||
2978 | if (clear_free_space_tree) { | |
f8d468a1 OS |
2979 | btrfs_info(fs_info, "clearing free space tree"); |
2980 | ret = btrfs_clear_free_space_tree(fs_info); | |
2981 | if (ret) { | |
2982 | btrfs_warn(fs_info, | |
2983 | "failed to clear free space tree: %d", ret); | |
6bccf3ab | 2984 | close_ctree(fs_info); |
f8d468a1 OS |
2985 | return ret; |
2986 | } | |
2987 | } | |
2988 | ||
0b246afa | 2989 | if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && |
511711af | 2990 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { |
05135f59 | 2991 | btrfs_info(fs_info, "creating free space tree"); |
511711af CM |
2992 | ret = btrfs_create_free_space_tree(fs_info); |
2993 | if (ret) { | |
05135f59 DS |
2994 | btrfs_warn(fs_info, |
2995 | "failed to create free space tree: %d", ret); | |
6bccf3ab | 2996 | close_ctree(fs_info); |
511711af CM |
2997 | return ret; |
2998 | } | |
2999 | } | |
3000 | ||
2b6ba629 ID |
3001 | down_read(&fs_info->cleanup_work_sem); |
3002 | if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || | |
3003 | (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { | |
e3acc2a6 | 3004 | up_read(&fs_info->cleanup_work_sem); |
6bccf3ab | 3005 | close_ctree(fs_info); |
2b6ba629 ID |
3006 | return ret; |
3007 | } | |
3008 | up_read(&fs_info->cleanup_work_sem); | |
59641015 | 3009 | |
2b6ba629 ID |
3010 | ret = btrfs_resume_balance_async(fs_info); |
3011 | if (ret) { | |
05135f59 | 3012 | btrfs_warn(fs_info, "failed to resume balance: %d", ret); |
6bccf3ab | 3013 | close_ctree(fs_info); |
2b6ba629 | 3014 | return ret; |
e3acc2a6 JB |
3015 | } |
3016 | ||
8dabb742 SB |
3017 | ret = btrfs_resume_dev_replace_async(fs_info); |
3018 | if (ret) { | |
05135f59 | 3019 | btrfs_warn(fs_info, "failed to resume device replace: %d", ret); |
6bccf3ab | 3020 | close_ctree(fs_info); |
8dabb742 SB |
3021 | return ret; |
3022 | } | |
3023 | ||
b382a324 JS |
3024 | btrfs_qgroup_rescan_resume(fs_info); |
3025 | ||
4bbcaa64 | 3026 | if (!fs_info->uuid_root) { |
05135f59 | 3027 | btrfs_info(fs_info, "creating UUID tree"); |
f7a81ea4 SB |
3028 | ret = btrfs_create_uuid_tree(fs_info); |
3029 | if (ret) { | |
05135f59 DS |
3030 | btrfs_warn(fs_info, |
3031 | "failed to create the UUID tree: %d", ret); | |
6bccf3ab | 3032 | close_ctree(fs_info); |
f7a81ea4 SB |
3033 | return ret; |
3034 | } | |
0b246afa | 3035 | } else if (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || |
4bbcaa64 ES |
3036 | fs_info->generation != |
3037 | btrfs_super_uuid_tree_generation(disk_super)) { | |
05135f59 | 3038 | btrfs_info(fs_info, "checking UUID tree"); |
70f80175 SB |
3039 | ret = btrfs_check_uuid_tree(fs_info); |
3040 | if (ret) { | |
05135f59 DS |
3041 | btrfs_warn(fs_info, |
3042 | "failed to check the UUID tree: %d", ret); | |
6bccf3ab | 3043 | close_ctree(fs_info); |
70f80175 SB |
3044 | return ret; |
3045 | } | |
3046 | } else { | |
afcdd129 | 3047 | set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); |
f7a81ea4 | 3048 | } |
afcdd129 | 3049 | set_bit(BTRFS_FS_OPEN, &fs_info->flags); |
47ab2a6c | 3050 | |
8dcddfa0 QW |
3051 | /* |
3052 | * backuproot only affect mount behavior, and if open_ctree succeeded, | |
3053 | * no need to keep the flag | |
3054 | */ | |
3055 | btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); | |
3056 | ||
ad2b2c80 | 3057 | return 0; |
39279cc3 | 3058 | |
bcef60f2 AJ |
3059 | fail_qgroup: |
3060 | btrfs_free_qgroup_config(fs_info); | |
7c2ca468 CM |
3061 | fail_trans_kthread: |
3062 | kthread_stop(fs_info->transaction_kthread); | |
2ff7e61e | 3063 | btrfs_cleanup_transaction(fs_info); |
faa2dbf0 | 3064 | btrfs_free_fs_roots(fs_info); |
3f157a2f | 3065 | fail_cleaner: |
a74a4b97 | 3066 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
3067 | |
3068 | /* | |
3069 | * make sure we're done with the btree inode before we stop our | |
3070 | * kthreads | |
3071 | */ | |
3072 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
7c2ca468 | 3073 | |
2365dd3c | 3074 | fail_sysfs: |
6618a59b | 3075 | btrfs_sysfs_remove_mounted(fs_info); |
2365dd3c | 3076 | |
b7c35e81 AJ |
3077 | fail_fsdev_sysfs: |
3078 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); | |
3079 | ||
1b1d1f66 | 3080 | fail_block_groups: |
54067ae9 | 3081 | btrfs_put_block_group_cache(fs_info); |
af31f5e5 CM |
3082 | |
3083 | fail_tree_roots: | |
3084 | free_root_pointers(fs_info, 1); | |
2b8195bb | 3085 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
af31f5e5 | 3086 | |
39279cc3 | 3087 | fail_sb_buffer: |
7abadb64 | 3088 | btrfs_stop_all_workers(fs_info); |
5cdd7db6 | 3089 | btrfs_free_block_groups(fs_info); |
16cdcec7 | 3090 | fail_alloc: |
4543df7e | 3091 | fail_iput: |
586e46e2 ID |
3092 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
3093 | ||
4543df7e | 3094 | iput(fs_info->btree_inode); |
c404e0dc MX |
3095 | fail_bio_counter: |
3096 | percpu_counter_destroy(&fs_info->bio_counter); | |
963d678b MX |
3097 | fail_delalloc_bytes: |
3098 | percpu_counter_destroy(&fs_info->delalloc_bytes); | |
e2d84521 MX |
3099 | fail_dirty_metadata_bytes: |
3100 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); | |
76dda93c YZ |
3101 | fail_srcu: |
3102 | cleanup_srcu_struct(&fs_info->subvol_srcu); | |
7e662854 | 3103 | fail: |
53b381b3 | 3104 | btrfs_free_stripe_hash_table(fs_info); |
586e46e2 | 3105 | btrfs_close_devices(fs_info->fs_devices); |
ad2b2c80 | 3106 | return err; |
af31f5e5 CM |
3107 | |
3108 | recovery_tree_root: | |
0b246afa | 3109 | if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) |
af31f5e5 CM |
3110 | goto fail_tree_roots; |
3111 | ||
3112 | free_root_pointers(fs_info, 0); | |
3113 | ||
3114 | /* don't use the log in recovery mode, it won't be valid */ | |
3115 | btrfs_set_super_log_root(disk_super, 0); | |
3116 | ||
3117 | /* we can't trust the free space cache either */ | |
3118 | btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); | |
3119 | ||
3120 | ret = next_root_backup(fs_info, fs_info->super_copy, | |
3121 | &num_backups_tried, &backup_index); | |
3122 | if (ret == -1) | |
3123 | goto fail_block_groups; | |
3124 | goto retry_root_backup; | |
eb60ceac CM |
3125 | } |
3126 | ||
f2984462 CM |
3127 | static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) |
3128 | { | |
f2984462 CM |
3129 | if (uptodate) { |
3130 | set_buffer_uptodate(bh); | |
3131 | } else { | |
442a4f63 SB |
3132 | struct btrfs_device *device = (struct btrfs_device *) |
3133 | bh->b_private; | |
3134 | ||
fb456252 | 3135 | btrfs_warn_rl_in_rcu(device->fs_info, |
b14af3b4 | 3136 | "lost page write due to IO error on %s", |
606686ee | 3137 | rcu_str_deref(device->name)); |
01327610 | 3138 | /* note, we don't set_buffer_write_io_error because we have |
1259ab75 CM |
3139 | * our own ways of dealing with the IO errors |
3140 | */ | |
f2984462 | 3141 | clear_buffer_uptodate(bh); |
442a4f63 | 3142 | btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS); |
f2984462 CM |
3143 | } |
3144 | unlock_buffer(bh); | |
3145 | put_bh(bh); | |
3146 | } | |
3147 | ||
29c36d72 AJ |
3148 | int btrfs_read_dev_one_super(struct block_device *bdev, int copy_num, |
3149 | struct buffer_head **bh_ret) | |
3150 | { | |
3151 | struct buffer_head *bh; | |
3152 | struct btrfs_super_block *super; | |
3153 | u64 bytenr; | |
3154 | ||
3155 | bytenr = btrfs_sb_offset(copy_num); | |
3156 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= i_size_read(bdev->bd_inode)) | |
3157 | return -EINVAL; | |
3158 | ||
9f6d2510 | 3159 | bh = __bread(bdev, bytenr / BTRFS_BDEV_BLOCKSIZE, BTRFS_SUPER_INFO_SIZE); |
29c36d72 AJ |
3160 | /* |
3161 | * If we fail to read from the underlying devices, as of now | |
3162 | * the best option we have is to mark it EIO. | |
3163 | */ | |
3164 | if (!bh) | |
3165 | return -EIO; | |
3166 | ||
3167 | super = (struct btrfs_super_block *)bh->b_data; | |
3168 | if (btrfs_super_bytenr(super) != bytenr || | |
3169 | btrfs_super_magic(super) != BTRFS_MAGIC) { | |
3170 | brelse(bh); | |
3171 | return -EINVAL; | |
3172 | } | |
3173 | ||
3174 | *bh_ret = bh; | |
3175 | return 0; | |
3176 | } | |
3177 | ||
3178 | ||
a512bbf8 YZ |
3179 | struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) |
3180 | { | |
3181 | struct buffer_head *bh; | |
3182 | struct buffer_head *latest = NULL; | |
3183 | struct btrfs_super_block *super; | |
3184 | int i; | |
3185 | u64 transid = 0; | |
92fc03fb | 3186 | int ret = -EINVAL; |
a512bbf8 YZ |
3187 | |
3188 | /* we would like to check all the supers, but that would make | |
3189 | * a btrfs mount succeed after a mkfs from a different FS. | |
3190 | * So, we need to add a special mount option to scan for | |
3191 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
3192 | */ | |
3193 | for (i = 0; i < 1; i++) { | |
29c36d72 AJ |
3194 | ret = btrfs_read_dev_one_super(bdev, i, &bh); |
3195 | if (ret) | |
a512bbf8 YZ |
3196 | continue; |
3197 | ||
3198 | super = (struct btrfs_super_block *)bh->b_data; | |
a512bbf8 YZ |
3199 | |
3200 | if (!latest || btrfs_super_generation(super) > transid) { | |
3201 | brelse(latest); | |
3202 | latest = bh; | |
3203 | transid = btrfs_super_generation(super); | |
3204 | } else { | |
3205 | brelse(bh); | |
3206 | } | |
3207 | } | |
92fc03fb AJ |
3208 | |
3209 | if (!latest) | |
3210 | return ERR_PTR(ret); | |
3211 | ||
a512bbf8 YZ |
3212 | return latest; |
3213 | } | |
3214 | ||
4eedeb75 | 3215 | /* |
abbb3b8e DS |
3216 | * Write superblock @sb to the @device. Do not wait for completion, all the |
3217 | * buffer heads we write are pinned. | |
4eedeb75 | 3218 | * |
abbb3b8e DS |
3219 | * Write @max_mirrors copies of the superblock, where 0 means default that fit |
3220 | * the expected device size at commit time. Note that max_mirrors must be | |
3221 | * same for write and wait phases. | |
4eedeb75 | 3222 | * |
abbb3b8e | 3223 | * Return number of errors when buffer head is not found or submission fails. |
4eedeb75 | 3224 | */ |
a512bbf8 | 3225 | static int write_dev_supers(struct btrfs_device *device, |
abbb3b8e | 3226 | struct btrfs_super_block *sb, int max_mirrors) |
a512bbf8 YZ |
3227 | { |
3228 | struct buffer_head *bh; | |
3229 | int i; | |
3230 | int ret; | |
3231 | int errors = 0; | |
3232 | u32 crc; | |
3233 | u64 bytenr; | |
a512bbf8 YZ |
3234 | |
3235 | if (max_mirrors == 0) | |
3236 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
3237 | ||
a512bbf8 YZ |
3238 | for (i = 0; i < max_mirrors; i++) { |
3239 | bytenr = btrfs_sb_offset(i); | |
935e5cc9 MX |
3240 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
3241 | device->commit_total_bytes) | |
a512bbf8 YZ |
3242 | break; |
3243 | ||
abbb3b8e | 3244 | btrfs_set_super_bytenr(sb, bytenr); |
4eedeb75 | 3245 | |
abbb3b8e DS |
3246 | crc = ~(u32)0; |
3247 | crc = btrfs_csum_data((const char *)sb + BTRFS_CSUM_SIZE, crc, | |
3248 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); | |
3249 | btrfs_csum_final(crc, sb->csum); | |
4eedeb75 | 3250 | |
abbb3b8e | 3251 | /* One reference for us, and we leave it for the caller */ |
9f6d2510 | 3252 | bh = __getblk(device->bdev, bytenr / BTRFS_BDEV_BLOCKSIZE, |
abbb3b8e DS |
3253 | BTRFS_SUPER_INFO_SIZE); |
3254 | if (!bh) { | |
3255 | btrfs_err(device->fs_info, | |
3256 | "couldn't get super buffer head for bytenr %llu", | |
3257 | bytenr); | |
3258 | errors++; | |
4eedeb75 | 3259 | continue; |
abbb3b8e | 3260 | } |
634554dc | 3261 | |
abbb3b8e | 3262 | memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); |
a512bbf8 | 3263 | |
abbb3b8e DS |
3264 | /* one reference for submit_bh */ |
3265 | get_bh(bh); | |
4eedeb75 | 3266 | |
abbb3b8e DS |
3267 | set_buffer_uptodate(bh); |
3268 | lock_buffer(bh); | |
3269 | bh->b_end_io = btrfs_end_buffer_write_sync; | |
3270 | bh->b_private = device; | |
a512bbf8 | 3271 | |
387125fc CM |
3272 | /* |
3273 | * we fua the first super. The others we allow | |
3274 | * to go down lazy. | |
3275 | */ | |
8d910125 JK |
3276 | if (i == 0) { |
3277 | ret = btrfsic_submit_bh(REQ_OP_WRITE, | |
db95c876 | 3278 | REQ_SYNC | REQ_FUA | REQ_META | REQ_PRIO, bh); |
8d910125 | 3279 | } else { |
db95c876 DS |
3280 | ret = btrfsic_submit_bh(REQ_OP_WRITE, |
3281 | REQ_SYNC | REQ_META | REQ_PRIO, bh); | |
8d910125 | 3282 | } |
4eedeb75 | 3283 | if (ret) |
a512bbf8 | 3284 | errors++; |
a512bbf8 YZ |
3285 | } |
3286 | return errors < i ? 0 : -1; | |
3287 | } | |
3288 | ||
abbb3b8e DS |
3289 | /* |
3290 | * Wait for write completion of superblocks done by write_dev_supers, | |
3291 | * @max_mirrors same for write and wait phases. | |
3292 | * | |
3293 | * Return number of errors when buffer head is not found or not marked up to | |
3294 | * date. | |
3295 | */ | |
3296 | static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) | |
3297 | { | |
3298 | struct buffer_head *bh; | |
3299 | int i; | |
3300 | int errors = 0; | |
3301 | u64 bytenr; | |
3302 | ||
3303 | if (max_mirrors == 0) | |
3304 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
3305 | ||
3306 | for (i = 0; i < max_mirrors; i++) { | |
3307 | bytenr = btrfs_sb_offset(i); | |
3308 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= | |
3309 | device->commit_total_bytes) | |
3310 | break; | |
3311 | ||
9f6d2510 DS |
3312 | bh = __find_get_block(device->bdev, |
3313 | bytenr / BTRFS_BDEV_BLOCKSIZE, | |
abbb3b8e DS |
3314 | BTRFS_SUPER_INFO_SIZE); |
3315 | if (!bh) { | |
3316 | errors++; | |
3317 | continue; | |
3318 | } | |
3319 | wait_on_buffer(bh); | |
3320 | if (!buffer_uptodate(bh)) | |
3321 | errors++; | |
3322 | ||
3323 | /* drop our reference */ | |
3324 | brelse(bh); | |
3325 | ||
3326 | /* drop the reference from the writing run */ | |
3327 | brelse(bh); | |
3328 | } | |
3329 | ||
3330 | return errors < i ? 0 : -1; | |
3331 | } | |
3332 | ||
387125fc CM |
3333 | /* |
3334 | * endio for the write_dev_flush, this will wake anyone waiting | |
3335 | * for the barrier when it is done | |
3336 | */ | |
4246a0b6 | 3337 | static void btrfs_end_empty_barrier(struct bio *bio) |
387125fc | 3338 | { |
e0ae9994 | 3339 | complete(bio->bi_private); |
387125fc CM |
3340 | } |
3341 | ||
3342 | /* | |
4fc6441a AJ |
3343 | * Submit a flush request to the device if it supports it. Error handling is |
3344 | * done in the waiting counterpart. | |
387125fc | 3345 | */ |
4fc6441a | 3346 | static void write_dev_flush(struct btrfs_device *device) |
387125fc | 3347 | { |
c2a9c7ab | 3348 | struct request_queue *q = bdev_get_queue(device->bdev); |
e0ae9994 | 3349 | struct bio *bio = device->flush_bio; |
387125fc | 3350 | |
c2a9c7ab | 3351 | if (!test_bit(QUEUE_FLAG_WC, &q->queue_flags)) |
4fc6441a | 3352 | return; |
387125fc | 3353 | |
e0ae9994 | 3354 | bio_reset(bio); |
387125fc | 3355 | bio->bi_end_io = btrfs_end_empty_barrier; |
74d46992 | 3356 | bio_set_dev(bio, device->bdev); |
8d910125 | 3357 | bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH; |
387125fc CM |
3358 | init_completion(&device->flush_wait); |
3359 | bio->bi_private = &device->flush_wait; | |
387125fc | 3360 | |
43a01111 | 3361 | btrfsic_submit_bio(bio); |
e0ae9994 | 3362 | device->flush_bio_sent = 1; |
4fc6441a | 3363 | } |
387125fc | 3364 | |
4fc6441a AJ |
3365 | /* |
3366 | * If the flush bio has been submitted by write_dev_flush, wait for it. | |
3367 | */ | |
8c27cb35 | 3368 | static blk_status_t wait_dev_flush(struct btrfs_device *device) |
4fc6441a | 3369 | { |
4fc6441a | 3370 | struct bio *bio = device->flush_bio; |
387125fc | 3371 | |
e0ae9994 | 3372 | if (!device->flush_bio_sent) |
58efbc9f | 3373 | return BLK_STS_OK; |
387125fc | 3374 | |
e0ae9994 | 3375 | device->flush_bio_sent = 0; |
2980d574 | 3376 | wait_for_completion_io(&device->flush_wait); |
387125fc | 3377 | |
8c27cb35 | 3378 | return bio->bi_status; |
387125fc | 3379 | } |
387125fc | 3380 | |
d10b82fe | 3381 | static int check_barrier_error(struct btrfs_fs_info *fs_info) |
401b41e5 | 3382 | { |
d10b82fe | 3383 | if (!btrfs_check_rw_degradable(fs_info)) |
401b41e5 | 3384 | return -EIO; |
387125fc CM |
3385 | return 0; |
3386 | } | |
3387 | ||
3388 | /* | |
3389 | * send an empty flush down to each device in parallel, | |
3390 | * then wait for them | |
3391 | */ | |
3392 | static int barrier_all_devices(struct btrfs_fs_info *info) | |
3393 | { | |
3394 | struct list_head *head; | |
3395 | struct btrfs_device *dev; | |
5af3e8cc | 3396 | int errors_wait = 0; |
4e4cbee9 | 3397 | blk_status_t ret; |
387125fc CM |
3398 | |
3399 | /* send down all the barriers */ | |
3400 | head = &info->fs_devices->devices; | |
3401 | list_for_each_entry_rcu(dev, head, dev_list) { | |
f88ba6a2 HS |
3402 | if (dev->missing) |
3403 | continue; | |
cea7c8bf | 3404 | if (!dev->bdev) |
387125fc | 3405 | continue; |
387125fc CM |
3406 | if (!dev->in_fs_metadata || !dev->writeable) |
3407 | continue; | |
3408 | ||
4fc6441a | 3409 | write_dev_flush(dev); |
58efbc9f | 3410 | dev->last_flush_error = BLK_STS_OK; |
387125fc CM |
3411 | } |
3412 | ||
3413 | /* wait for all the barriers */ | |
3414 | list_for_each_entry_rcu(dev, head, dev_list) { | |
f88ba6a2 HS |
3415 | if (dev->missing) |
3416 | continue; | |
387125fc | 3417 | if (!dev->bdev) { |
5af3e8cc | 3418 | errors_wait++; |
387125fc CM |
3419 | continue; |
3420 | } | |
3421 | if (!dev->in_fs_metadata || !dev->writeable) | |
3422 | continue; | |
3423 | ||
4fc6441a | 3424 | ret = wait_dev_flush(dev); |
401b41e5 AJ |
3425 | if (ret) { |
3426 | dev->last_flush_error = ret; | |
66b4993e DS |
3427 | btrfs_dev_stat_inc_and_print(dev, |
3428 | BTRFS_DEV_STAT_FLUSH_ERRS); | |
5af3e8cc | 3429 | errors_wait++; |
401b41e5 AJ |
3430 | } |
3431 | } | |
3432 | ||
cea7c8bf | 3433 | if (errors_wait) { |
401b41e5 AJ |
3434 | /* |
3435 | * At some point we need the status of all disks | |
3436 | * to arrive at the volume status. So error checking | |
3437 | * is being pushed to a separate loop. | |
3438 | */ | |
d10b82fe | 3439 | return check_barrier_error(info); |
387125fc | 3440 | } |
387125fc CM |
3441 | return 0; |
3442 | } | |
3443 | ||
943c6e99 ZL |
3444 | int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) |
3445 | { | |
8789f4fe ZL |
3446 | int raid_type; |
3447 | int min_tolerated = INT_MAX; | |
943c6e99 | 3448 | |
8789f4fe ZL |
3449 | if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || |
3450 | (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) | |
3451 | min_tolerated = min(min_tolerated, | |
3452 | btrfs_raid_array[BTRFS_RAID_SINGLE]. | |
3453 | tolerated_failures); | |
943c6e99 | 3454 | |
8789f4fe ZL |
3455 | for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { |
3456 | if (raid_type == BTRFS_RAID_SINGLE) | |
3457 | continue; | |
3458 | if (!(flags & btrfs_raid_group[raid_type])) | |
3459 | continue; | |
3460 | min_tolerated = min(min_tolerated, | |
3461 | btrfs_raid_array[raid_type]. | |
3462 | tolerated_failures); | |
3463 | } | |
943c6e99 | 3464 | |
8789f4fe | 3465 | if (min_tolerated == INT_MAX) { |
ab8d0fc4 | 3466 | pr_warn("BTRFS: unknown raid flag: %llu", flags); |
8789f4fe ZL |
3467 | min_tolerated = 0; |
3468 | } | |
3469 | ||
3470 | return min_tolerated; | |
943c6e99 ZL |
3471 | } |
3472 | ||
eece6a9c | 3473 | int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) |
f2984462 | 3474 | { |
e5e9a520 | 3475 | struct list_head *head; |
f2984462 | 3476 | struct btrfs_device *dev; |
a061fc8d | 3477 | struct btrfs_super_block *sb; |
f2984462 | 3478 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
3479 | int ret; |
3480 | int do_barriers; | |
a236aed1 CM |
3481 | int max_errors; |
3482 | int total_errors = 0; | |
a061fc8d | 3483 | u64 flags; |
f2984462 | 3484 | |
0b246afa | 3485 | do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); |
fed3b381 LB |
3486 | |
3487 | /* | |
3488 | * max_mirrors == 0 indicates we're from commit_transaction, | |
3489 | * not from fsync where the tree roots in fs_info have not | |
3490 | * been consistent on disk. | |
3491 | */ | |
3492 | if (max_mirrors == 0) | |
3493 | backup_super_roots(fs_info); | |
f2984462 | 3494 | |
0b246afa | 3495 | sb = fs_info->super_for_commit; |
a061fc8d | 3496 | dev_item = &sb->dev_item; |
e5e9a520 | 3497 | |
0b246afa JM |
3498 | mutex_lock(&fs_info->fs_devices->device_list_mutex); |
3499 | head = &fs_info->fs_devices->devices; | |
3500 | max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; | |
387125fc | 3501 | |
5af3e8cc | 3502 | if (do_barriers) { |
0b246afa | 3503 | ret = barrier_all_devices(fs_info); |
5af3e8cc SB |
3504 | if (ret) { |
3505 | mutex_unlock( | |
0b246afa JM |
3506 | &fs_info->fs_devices->device_list_mutex); |
3507 | btrfs_handle_fs_error(fs_info, ret, | |
3508 | "errors while submitting device barriers."); | |
5af3e8cc SB |
3509 | return ret; |
3510 | } | |
3511 | } | |
387125fc | 3512 | |
1f78160c | 3513 | list_for_each_entry_rcu(dev, head, dev_list) { |
dfe25020 CM |
3514 | if (!dev->bdev) { |
3515 | total_errors++; | |
3516 | continue; | |
3517 | } | |
2b82032c | 3518 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
3519 | continue; |
3520 | ||
2b82032c | 3521 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
3522 | btrfs_set_stack_device_type(dev_item, dev->type); |
3523 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
7df69d3e | 3524 | btrfs_set_stack_device_total_bytes(dev_item, |
935e5cc9 | 3525 | dev->commit_total_bytes); |
ce7213c7 MX |
3526 | btrfs_set_stack_device_bytes_used(dev_item, |
3527 | dev->commit_bytes_used); | |
a061fc8d CM |
3528 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); |
3529 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
3530 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
3531 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
44880fdc | 3532 | memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_FSID_SIZE); |
a512bbf8 | 3533 | |
a061fc8d CM |
3534 | flags = btrfs_super_flags(sb); |
3535 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
3536 | ||
abbb3b8e | 3537 | ret = write_dev_supers(dev, sb, max_mirrors); |
a236aed1 CM |
3538 | if (ret) |
3539 | total_errors++; | |
f2984462 | 3540 | } |
a236aed1 | 3541 | if (total_errors > max_errors) { |
0b246afa JM |
3542 | btrfs_err(fs_info, "%d errors while writing supers", |
3543 | total_errors); | |
3544 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
79787eaa | 3545 | |
9d565ba4 | 3546 | /* FUA is masked off if unsupported and can't be the reason */ |
0b246afa JM |
3547 | btrfs_handle_fs_error(fs_info, -EIO, |
3548 | "%d errors while writing supers", | |
3549 | total_errors); | |
9d565ba4 | 3550 | return -EIO; |
a236aed1 | 3551 | } |
f2984462 | 3552 | |
a512bbf8 | 3553 | total_errors = 0; |
1f78160c | 3554 | list_for_each_entry_rcu(dev, head, dev_list) { |
dfe25020 CM |
3555 | if (!dev->bdev) |
3556 | continue; | |
2b82032c | 3557 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
3558 | continue; |
3559 | ||
abbb3b8e | 3560 | ret = wait_dev_supers(dev, max_mirrors); |
a512bbf8 YZ |
3561 | if (ret) |
3562 | total_errors++; | |
f2984462 | 3563 | } |
0b246afa | 3564 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); |
a236aed1 | 3565 | if (total_errors > max_errors) { |
0b246afa JM |
3566 | btrfs_handle_fs_error(fs_info, -EIO, |
3567 | "%d errors while writing supers", | |
3568 | total_errors); | |
79787eaa | 3569 | return -EIO; |
a236aed1 | 3570 | } |
f2984462 CM |
3571 | return 0; |
3572 | } | |
3573 | ||
cb517eab MX |
3574 | /* Drop a fs root from the radix tree and free it. */ |
3575 | void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, | |
3576 | struct btrfs_root *root) | |
2619ba1f | 3577 | { |
4df27c4d | 3578 | spin_lock(&fs_info->fs_roots_radix_lock); |
2619ba1f CM |
3579 | radix_tree_delete(&fs_info->fs_roots_radix, |
3580 | (unsigned long)root->root_key.objectid); | |
4df27c4d | 3581 | spin_unlock(&fs_info->fs_roots_radix_lock); |
76dda93c YZ |
3582 | |
3583 | if (btrfs_root_refs(&root->root_item) == 0) | |
3584 | synchronize_srcu(&fs_info->subvol_srcu); | |
3585 | ||
1c1ea4f7 | 3586 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { |
3321719e | 3587 | btrfs_free_log(NULL, root); |
1c1ea4f7 LB |
3588 | if (root->reloc_root) { |
3589 | free_extent_buffer(root->reloc_root->node); | |
3590 | free_extent_buffer(root->reloc_root->commit_root); | |
3591 | btrfs_put_fs_root(root->reloc_root); | |
3592 | root->reloc_root = NULL; | |
3593 | } | |
3594 | } | |
3321719e | 3595 | |
faa2dbf0 JB |
3596 | if (root->free_ino_pinned) |
3597 | __btrfs_remove_free_space_cache(root->free_ino_pinned); | |
3598 | if (root->free_ino_ctl) | |
3599 | __btrfs_remove_free_space_cache(root->free_ino_ctl); | |
4df27c4d | 3600 | free_fs_root(root); |
4df27c4d YZ |
3601 | } |
3602 | ||
3603 | static void free_fs_root(struct btrfs_root *root) | |
3604 | { | |
57cdc8db | 3605 | iput(root->ino_cache_inode); |
4df27c4d | 3606 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); |
2ff7e61e | 3607 | btrfs_free_block_rsv(root->fs_info, root->orphan_block_rsv); |
1cb048f5 | 3608 | root->orphan_block_rsv = NULL; |
0ee5dc67 AV |
3609 | if (root->anon_dev) |
3610 | free_anon_bdev(root->anon_dev); | |
8257b2dc MX |
3611 | if (root->subv_writers) |
3612 | btrfs_free_subvolume_writers(root->subv_writers); | |
4df27c4d YZ |
3613 | free_extent_buffer(root->node); |
3614 | free_extent_buffer(root->commit_root); | |
581bb050 LZ |
3615 | kfree(root->free_ino_ctl); |
3616 | kfree(root->free_ino_pinned); | |
d397712b | 3617 | kfree(root->name); |
b0feb9d9 | 3618 | btrfs_put_fs_root(root); |
2619ba1f CM |
3619 | } |
3620 | ||
cb517eab MX |
3621 | void btrfs_free_fs_root(struct btrfs_root *root) |
3622 | { | |
3623 | free_fs_root(root); | |
2619ba1f CM |
3624 | } |
3625 | ||
c146afad | 3626 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 3627 | { |
c146afad YZ |
3628 | u64 root_objectid = 0; |
3629 | struct btrfs_root *gang[8]; | |
65d33fd7 QW |
3630 | int i = 0; |
3631 | int err = 0; | |
3632 | unsigned int ret = 0; | |
3633 | int index; | |
e089f05c | 3634 | |
c146afad | 3635 | while (1) { |
65d33fd7 | 3636 | index = srcu_read_lock(&fs_info->subvol_srcu); |
c146afad YZ |
3637 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, |
3638 | (void **)gang, root_objectid, | |
3639 | ARRAY_SIZE(gang)); | |
65d33fd7 QW |
3640 | if (!ret) { |
3641 | srcu_read_unlock(&fs_info->subvol_srcu, index); | |
c146afad | 3642 | break; |
65d33fd7 | 3643 | } |
5d4f98a2 | 3644 | root_objectid = gang[ret - 1]->root_key.objectid + 1; |
65d33fd7 | 3645 | |
c146afad | 3646 | for (i = 0; i < ret; i++) { |
65d33fd7 QW |
3647 | /* Avoid to grab roots in dead_roots */ |
3648 | if (btrfs_root_refs(&gang[i]->root_item) == 0) { | |
3649 | gang[i] = NULL; | |
3650 | continue; | |
3651 | } | |
3652 | /* grab all the search result for later use */ | |
3653 | gang[i] = btrfs_grab_fs_root(gang[i]); | |
3654 | } | |
3655 | srcu_read_unlock(&fs_info->subvol_srcu, index); | |
66b4ffd1 | 3656 | |
65d33fd7 QW |
3657 | for (i = 0; i < ret; i++) { |
3658 | if (!gang[i]) | |
3659 | continue; | |
c146afad | 3660 | root_objectid = gang[i]->root_key.objectid; |
66b4ffd1 JB |
3661 | err = btrfs_orphan_cleanup(gang[i]); |
3662 | if (err) | |
65d33fd7 QW |
3663 | break; |
3664 | btrfs_put_fs_root(gang[i]); | |
c146afad YZ |
3665 | } |
3666 | root_objectid++; | |
3667 | } | |
65d33fd7 QW |
3668 | |
3669 | /* release the uncleaned roots due to error */ | |
3670 | for (; i < ret; i++) { | |
3671 | if (gang[i]) | |
3672 | btrfs_put_fs_root(gang[i]); | |
3673 | } | |
3674 | return err; | |
c146afad | 3675 | } |
a2135011 | 3676 | |
6bccf3ab | 3677 | int btrfs_commit_super(struct btrfs_fs_info *fs_info) |
c146afad | 3678 | { |
6bccf3ab | 3679 | struct btrfs_root *root = fs_info->tree_root; |
c146afad | 3680 | struct btrfs_trans_handle *trans; |
a74a4b97 | 3681 | |
0b246afa | 3682 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 3683 | btrfs_run_delayed_iputs(fs_info); |
0b246afa JM |
3684 | mutex_unlock(&fs_info->cleaner_mutex); |
3685 | wake_up_process(fs_info->cleaner_kthread); | |
c71bf099 YZ |
3686 | |
3687 | /* wait until ongoing cleanup work done */ | |
0b246afa JM |
3688 | down_write(&fs_info->cleanup_work_sem); |
3689 | up_write(&fs_info->cleanup_work_sem); | |
c71bf099 | 3690 | |
7a7eaa40 | 3691 | trans = btrfs_join_transaction(root); |
3612b495 TI |
3692 | if (IS_ERR(trans)) |
3693 | return PTR_ERR(trans); | |
3a45bb20 | 3694 | return btrfs_commit_transaction(trans); |
c146afad YZ |
3695 | } |
3696 | ||
6bccf3ab | 3697 | void close_ctree(struct btrfs_fs_info *fs_info) |
c146afad | 3698 | { |
6bccf3ab | 3699 | struct btrfs_root *root = fs_info->tree_root; |
c146afad YZ |
3700 | int ret; |
3701 | ||
afcdd129 | 3702 | set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); |
c146afad | 3703 | |
7343dd61 | 3704 | /* wait for the qgroup rescan worker to stop */ |
d06f23d6 | 3705 | btrfs_qgroup_wait_for_completion(fs_info, false); |
7343dd61 | 3706 | |
803b2f54 SB |
3707 | /* wait for the uuid_scan task to finish */ |
3708 | down(&fs_info->uuid_tree_rescan_sem); | |
3709 | /* avoid complains from lockdep et al., set sem back to initial state */ | |
3710 | up(&fs_info->uuid_tree_rescan_sem); | |
3711 | ||
837d5b6e | 3712 | /* pause restriper - we want to resume on mount */ |
aa1b8cd4 | 3713 | btrfs_pause_balance(fs_info); |
837d5b6e | 3714 | |
8dabb742 SB |
3715 | btrfs_dev_replace_suspend_for_unmount(fs_info); |
3716 | ||
aa1b8cd4 | 3717 | btrfs_scrub_cancel(fs_info); |
4cb5300b CM |
3718 | |
3719 | /* wait for any defraggers to finish */ | |
3720 | wait_event(fs_info->transaction_wait, | |
3721 | (atomic_read(&fs_info->defrag_running) == 0)); | |
3722 | ||
3723 | /* clear out the rbtree of defraggable inodes */ | |
26176e7c | 3724 | btrfs_cleanup_defrag_inodes(fs_info); |
4cb5300b | 3725 | |
21c7e756 MX |
3726 | cancel_work_sync(&fs_info->async_reclaim_work); |
3727 | ||
bc98a42c | 3728 | if (!sb_rdonly(fs_info->sb)) { |
e44163e1 JM |
3729 | /* |
3730 | * If the cleaner thread is stopped and there are | |
3731 | * block groups queued for removal, the deletion will be | |
3732 | * skipped when we quit the cleaner thread. | |
3733 | */ | |
0b246afa | 3734 | btrfs_delete_unused_bgs(fs_info); |
e44163e1 | 3735 | |
6bccf3ab | 3736 | ret = btrfs_commit_super(fs_info); |
acce952b | 3737 | if (ret) |
04892340 | 3738 | btrfs_err(fs_info, "commit super ret %d", ret); |
acce952b | 3739 | } |
3740 | ||
87533c47 | 3741 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) |
2ff7e61e | 3742 | btrfs_error_commit_super(fs_info); |
0f7d52f4 | 3743 | |
e3029d9f AV |
3744 | kthread_stop(fs_info->transaction_kthread); |
3745 | kthread_stop(fs_info->cleaner_kthread); | |
8929ecfa | 3746 | |
afcdd129 | 3747 | set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); |
f25784b3 | 3748 | |
04892340 | 3749 | btrfs_free_qgroup_config(fs_info); |
bcef60f2 | 3750 | |
963d678b | 3751 | if (percpu_counter_sum(&fs_info->delalloc_bytes)) { |
04892340 | 3752 | btrfs_info(fs_info, "at unmount delalloc count %lld", |
963d678b | 3753 | percpu_counter_sum(&fs_info->delalloc_bytes)); |
b0c68f8b | 3754 | } |
bcc63abb | 3755 | |
6618a59b | 3756 | btrfs_sysfs_remove_mounted(fs_info); |
b7c35e81 | 3757 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); |
5ac1d209 | 3758 | |
faa2dbf0 | 3759 | btrfs_free_fs_roots(fs_info); |
d10c5f31 | 3760 | |
1a4319cc LB |
3761 | btrfs_put_block_group_cache(fs_info); |
3762 | ||
de348ee0 WS |
3763 | /* |
3764 | * we must make sure there is not any read request to | |
3765 | * submit after we stopping all workers. | |
3766 | */ | |
3767 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
96192499 JB |
3768 | btrfs_stop_all_workers(fs_info); |
3769 | ||
5cdd7db6 FM |
3770 | btrfs_free_block_groups(fs_info); |
3771 | ||
afcdd129 | 3772 | clear_bit(BTRFS_FS_OPEN, &fs_info->flags); |
13e6c37b | 3773 | free_root_pointers(fs_info, 1); |
9ad6b7bc | 3774 | |
13e6c37b | 3775 | iput(fs_info->btree_inode); |
d6bfde87 | 3776 | |
21adbd5c | 3777 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
0b246afa | 3778 | if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) |
2ff7e61e | 3779 | btrfsic_unmount(fs_info->fs_devices); |
21adbd5c SB |
3780 | #endif |
3781 | ||
dfe25020 | 3782 | btrfs_close_devices(fs_info->fs_devices); |
0b86a832 | 3783 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
b248a415 | 3784 | |
e2d84521 | 3785 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); |
963d678b | 3786 | percpu_counter_destroy(&fs_info->delalloc_bytes); |
c404e0dc | 3787 | percpu_counter_destroy(&fs_info->bio_counter); |
76dda93c | 3788 | cleanup_srcu_struct(&fs_info->subvol_srcu); |
0b86a832 | 3789 | |
53b381b3 | 3790 | btrfs_free_stripe_hash_table(fs_info); |
fd708b81 | 3791 | btrfs_free_ref_cache(fs_info); |
53b381b3 | 3792 | |
cdfb080e | 3793 | __btrfs_free_block_rsv(root->orphan_block_rsv); |
1cb048f5 | 3794 | root->orphan_block_rsv = NULL; |
04216820 | 3795 | |
04216820 FM |
3796 | while (!list_empty(&fs_info->pinned_chunks)) { |
3797 | struct extent_map *em; | |
3798 | ||
3799 | em = list_first_entry(&fs_info->pinned_chunks, | |
3800 | struct extent_map, list); | |
3801 | list_del_init(&em->list); | |
3802 | free_extent_map(em); | |
3803 | } | |
eb60ceac CM |
3804 | } |
3805 | ||
b9fab919 CM |
3806 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, |
3807 | int atomic) | |
5f39d397 | 3808 | { |
1259ab75 | 3809 | int ret; |
727011e0 | 3810 | struct inode *btree_inode = buf->pages[0]->mapping->host; |
1259ab75 | 3811 | |
0b32f4bb | 3812 | ret = extent_buffer_uptodate(buf); |
1259ab75 CM |
3813 | if (!ret) |
3814 | return ret; | |
3815 | ||
3816 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
b9fab919 CM |
3817 | parent_transid, atomic); |
3818 | if (ret == -EAGAIN) | |
3819 | return ret; | |
1259ab75 | 3820 | return !ret; |
5f39d397 CM |
3821 | } |
3822 | ||
5f39d397 CM |
3823 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
3824 | { | |
0b246afa | 3825 | struct btrfs_fs_info *fs_info; |
06ea65a3 | 3826 | struct btrfs_root *root; |
5f39d397 | 3827 | u64 transid = btrfs_header_generation(buf); |
b9473439 | 3828 | int was_dirty; |
b4ce94de | 3829 | |
06ea65a3 JB |
3830 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
3831 | /* | |
3832 | * This is a fast path so only do this check if we have sanity tests | |
3833 | * enabled. Normal people shouldn't be marking dummy buffers as dirty | |
3834 | * outside of the sanity tests. | |
3835 | */ | |
3836 | if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags))) | |
3837 | return; | |
3838 | #endif | |
3839 | root = BTRFS_I(buf->pages[0]->mapping->host)->root; | |
0b246afa | 3840 | fs_info = root->fs_info; |
b9447ef8 | 3841 | btrfs_assert_tree_locked(buf); |
0b246afa | 3842 | if (transid != fs_info->generation) |
5d163e0e | 3843 | WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", |
0b246afa | 3844 | buf->start, transid, fs_info->generation); |
0b32f4bb | 3845 | was_dirty = set_extent_buffer_dirty(buf); |
e2d84521 | 3846 | if (!was_dirty) |
104b4e51 NB |
3847 | percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, |
3848 | buf->len, | |
3849 | fs_info->dirty_metadata_batch); | |
1f21ef0a | 3850 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
557ea5dd | 3851 | if (btrfs_header_level(buf) == 0 && btrfs_check_leaf(root, buf)) { |
a4f78750 | 3852 | btrfs_print_leaf(buf); |
1f21ef0a FM |
3853 | ASSERT(0); |
3854 | } | |
3855 | #endif | |
eb60ceac CM |
3856 | } |
3857 | ||
2ff7e61e | 3858 | static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, |
b53d3f5d | 3859 | int flush_delayed) |
16cdcec7 MX |
3860 | { |
3861 | /* | |
3862 | * looks as though older kernels can get into trouble with | |
3863 | * this code, they end up stuck in balance_dirty_pages forever | |
3864 | */ | |
e2d84521 | 3865 | int ret; |
16cdcec7 MX |
3866 | |
3867 | if (current->flags & PF_MEMALLOC) | |
3868 | return; | |
3869 | ||
b53d3f5d | 3870 | if (flush_delayed) |
2ff7e61e | 3871 | btrfs_balance_delayed_items(fs_info); |
16cdcec7 | 3872 | |
0b246afa | 3873 | ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
e2d84521 MX |
3874 | BTRFS_DIRTY_METADATA_THRESH); |
3875 | if (ret > 0) { | |
0b246afa | 3876 | balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); |
16cdcec7 | 3877 | } |
16cdcec7 MX |
3878 | } |
3879 | ||
2ff7e61e | 3880 | void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) |
35b7e476 | 3881 | { |
2ff7e61e | 3882 | __btrfs_btree_balance_dirty(fs_info, 1); |
b53d3f5d | 3883 | } |
585ad2c3 | 3884 | |
2ff7e61e | 3885 | void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) |
b53d3f5d | 3886 | { |
2ff7e61e | 3887 | __btrfs_btree_balance_dirty(fs_info, 0); |
35b7e476 | 3888 | } |
6b80053d | 3889 | |
ca7a79ad | 3890 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) |
6b80053d | 3891 | { |
727011e0 | 3892 | struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; |
2ff7e61e JM |
3893 | struct btrfs_fs_info *fs_info = root->fs_info; |
3894 | ||
3895 | return btree_read_extent_buffer_pages(fs_info, buf, parent_transid); | |
6b80053d | 3896 | } |
0da5468f | 3897 | |
3d3a126a | 3898 | static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info) |
acce952b | 3899 | { |
c926093e | 3900 | struct btrfs_super_block *sb = fs_info->super_copy; |
319e4d06 QW |
3901 | u64 nodesize = btrfs_super_nodesize(sb); |
3902 | u64 sectorsize = btrfs_super_sectorsize(sb); | |
c926093e DS |
3903 | int ret = 0; |
3904 | ||
319e4d06 | 3905 | if (btrfs_super_magic(sb) != BTRFS_MAGIC) { |
ab8d0fc4 | 3906 | btrfs_err(fs_info, "no valid FS found"); |
319e4d06 QW |
3907 | ret = -EINVAL; |
3908 | } | |
3909 | if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) | |
ab8d0fc4 | 3910 | btrfs_warn(fs_info, "unrecognized super flag: %llu", |
319e4d06 | 3911 | btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); |
21e7626b | 3912 | if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { |
ab8d0fc4 | 3913 | btrfs_err(fs_info, "tree_root level too big: %d >= %d", |
21e7626b | 3914 | btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); |
c926093e DS |
3915 | ret = -EINVAL; |
3916 | } | |
21e7626b | 3917 | if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { |
ab8d0fc4 | 3918 | btrfs_err(fs_info, "chunk_root level too big: %d >= %d", |
21e7626b | 3919 | btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); |
c926093e DS |
3920 | ret = -EINVAL; |
3921 | } | |
21e7626b | 3922 | if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { |
ab8d0fc4 | 3923 | btrfs_err(fs_info, "log_root level too big: %d >= %d", |
21e7626b | 3924 | btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); |
c926093e DS |
3925 | ret = -EINVAL; |
3926 | } | |
3927 | ||
1104a885 | 3928 | /* |
319e4d06 QW |
3929 | * Check sectorsize and nodesize first, other check will need it. |
3930 | * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. | |
1104a885 | 3931 | */ |
319e4d06 QW |
3932 | if (!is_power_of_2(sectorsize) || sectorsize < 4096 || |
3933 | sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
ab8d0fc4 | 3934 | btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); |
319e4d06 QW |
3935 | ret = -EINVAL; |
3936 | } | |
3937 | /* Only PAGE SIZE is supported yet */ | |
09cbfeaf | 3938 | if (sectorsize != PAGE_SIZE) { |
ab8d0fc4 JM |
3939 | btrfs_err(fs_info, |
3940 | "sectorsize %llu not supported yet, only support %lu", | |
3941 | sectorsize, PAGE_SIZE); | |
319e4d06 QW |
3942 | ret = -EINVAL; |
3943 | } | |
3944 | if (!is_power_of_2(nodesize) || nodesize < sectorsize || | |
3945 | nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
ab8d0fc4 | 3946 | btrfs_err(fs_info, "invalid nodesize %llu", nodesize); |
319e4d06 QW |
3947 | ret = -EINVAL; |
3948 | } | |
3949 | if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { | |
ab8d0fc4 JM |
3950 | btrfs_err(fs_info, "invalid leafsize %u, should be %llu", |
3951 | le32_to_cpu(sb->__unused_leafsize), nodesize); | |
319e4d06 QW |
3952 | ret = -EINVAL; |
3953 | } | |
3954 | ||
3955 | /* Root alignment check */ | |
3956 | if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { | |
ab8d0fc4 JM |
3957 | btrfs_warn(fs_info, "tree_root block unaligned: %llu", |
3958 | btrfs_super_root(sb)); | |
319e4d06 QW |
3959 | ret = -EINVAL; |
3960 | } | |
3961 | if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { | |
ab8d0fc4 JM |
3962 | btrfs_warn(fs_info, "chunk_root block unaligned: %llu", |
3963 | btrfs_super_chunk_root(sb)); | |
75d6ad38 DS |
3964 | ret = -EINVAL; |
3965 | } | |
319e4d06 | 3966 | if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { |
ab8d0fc4 JM |
3967 | btrfs_warn(fs_info, "log_root block unaligned: %llu", |
3968 | btrfs_super_log_root(sb)); | |
75d6ad38 DS |
3969 | ret = -EINVAL; |
3970 | } | |
3971 | ||
44880fdc | 3972 | if (memcmp(fs_info->fsid, sb->dev_item.fsid, BTRFS_FSID_SIZE) != 0) { |
ab8d0fc4 JM |
3973 | btrfs_err(fs_info, |
3974 | "dev_item UUID does not match fsid: %pU != %pU", | |
3975 | fs_info->fsid, sb->dev_item.fsid); | |
c926093e DS |
3976 | ret = -EINVAL; |
3977 | } | |
3978 | ||
3979 | /* | |
3980 | * Hint to catch really bogus numbers, bitflips or so, more exact checks are | |
3981 | * done later | |
3982 | */ | |
99e3ecfc LB |
3983 | if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { |
3984 | btrfs_err(fs_info, "bytes_used is too small %llu", | |
ab8d0fc4 | 3985 | btrfs_super_bytes_used(sb)); |
99e3ecfc LB |
3986 | ret = -EINVAL; |
3987 | } | |
b7f67055 | 3988 | if (!is_power_of_2(btrfs_super_stripesize(sb))) { |
99e3ecfc | 3989 | btrfs_err(fs_info, "invalid stripesize %u", |
ab8d0fc4 | 3990 | btrfs_super_stripesize(sb)); |
99e3ecfc LB |
3991 | ret = -EINVAL; |
3992 | } | |
21e7626b | 3993 | if (btrfs_super_num_devices(sb) > (1UL << 31)) |
ab8d0fc4 JM |
3994 | btrfs_warn(fs_info, "suspicious number of devices: %llu", |
3995 | btrfs_super_num_devices(sb)); | |
75d6ad38 | 3996 | if (btrfs_super_num_devices(sb) == 0) { |
ab8d0fc4 | 3997 | btrfs_err(fs_info, "number of devices is 0"); |
75d6ad38 DS |
3998 | ret = -EINVAL; |
3999 | } | |
c926093e | 4000 | |
21e7626b | 4001 | if (btrfs_super_bytenr(sb) != BTRFS_SUPER_INFO_OFFSET) { |
ab8d0fc4 JM |
4002 | btrfs_err(fs_info, "super offset mismatch %llu != %u", |
4003 | btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); | |
c926093e DS |
4004 | ret = -EINVAL; |
4005 | } | |
4006 | ||
ce7fca5f DS |
4007 | /* |
4008 | * Obvious sys_chunk_array corruptions, it must hold at least one key | |
4009 | * and one chunk | |
4010 | */ | |
4011 | if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { | |
ab8d0fc4 JM |
4012 | btrfs_err(fs_info, "system chunk array too big %u > %u", |
4013 | btrfs_super_sys_array_size(sb), | |
4014 | BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); | |
ce7fca5f DS |
4015 | ret = -EINVAL; |
4016 | } | |
4017 | if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) | |
4018 | + sizeof(struct btrfs_chunk)) { | |
ab8d0fc4 JM |
4019 | btrfs_err(fs_info, "system chunk array too small %u < %zu", |
4020 | btrfs_super_sys_array_size(sb), | |
4021 | sizeof(struct btrfs_disk_key) | |
4022 | + sizeof(struct btrfs_chunk)); | |
ce7fca5f DS |
4023 | ret = -EINVAL; |
4024 | } | |
4025 | ||
c926093e DS |
4026 | /* |
4027 | * The generation is a global counter, we'll trust it more than the others | |
4028 | * but it's still possible that it's the one that's wrong. | |
4029 | */ | |
21e7626b | 4030 | if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) |
ab8d0fc4 JM |
4031 | btrfs_warn(fs_info, |
4032 | "suspicious: generation < chunk_root_generation: %llu < %llu", | |
4033 | btrfs_super_generation(sb), | |
4034 | btrfs_super_chunk_root_generation(sb)); | |
21e7626b DS |
4035 | if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) |
4036 | && btrfs_super_cache_generation(sb) != (u64)-1) | |
ab8d0fc4 JM |
4037 | btrfs_warn(fs_info, |
4038 | "suspicious: generation < cache_generation: %llu < %llu", | |
4039 | btrfs_super_generation(sb), | |
4040 | btrfs_super_cache_generation(sb)); | |
c926093e DS |
4041 | |
4042 | return ret; | |
acce952b | 4043 | } |
4044 | ||
2ff7e61e | 4045 | static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) |
acce952b | 4046 | { |
0b246afa | 4047 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 4048 | btrfs_run_delayed_iputs(fs_info); |
0b246afa | 4049 | mutex_unlock(&fs_info->cleaner_mutex); |
acce952b | 4050 | |
0b246afa JM |
4051 | down_write(&fs_info->cleanup_work_sem); |
4052 | up_write(&fs_info->cleanup_work_sem); | |
acce952b | 4053 | |
4054 | /* cleanup FS via transaction */ | |
2ff7e61e | 4055 | btrfs_cleanup_transaction(fs_info); |
acce952b | 4056 | } |
4057 | ||
143bede5 | 4058 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root) |
acce952b | 4059 | { |
acce952b | 4060 | struct btrfs_ordered_extent *ordered; |
acce952b | 4061 | |
199c2a9c | 4062 | spin_lock(&root->ordered_extent_lock); |
779880ef JB |
4063 | /* |
4064 | * This will just short circuit the ordered completion stuff which will | |
4065 | * make sure the ordered extent gets properly cleaned up. | |
4066 | */ | |
199c2a9c | 4067 | list_for_each_entry(ordered, &root->ordered_extents, |
779880ef JB |
4068 | root_extent_list) |
4069 | set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); | |
199c2a9c MX |
4070 | spin_unlock(&root->ordered_extent_lock); |
4071 | } | |
4072 | ||
4073 | static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) | |
4074 | { | |
4075 | struct btrfs_root *root; | |
4076 | struct list_head splice; | |
4077 | ||
4078 | INIT_LIST_HEAD(&splice); | |
4079 | ||
4080 | spin_lock(&fs_info->ordered_root_lock); | |
4081 | list_splice_init(&fs_info->ordered_roots, &splice); | |
4082 | while (!list_empty(&splice)) { | |
4083 | root = list_first_entry(&splice, struct btrfs_root, | |
4084 | ordered_root); | |
1de2cfde JB |
4085 | list_move_tail(&root->ordered_root, |
4086 | &fs_info->ordered_roots); | |
199c2a9c | 4087 | |
2a85d9ca | 4088 | spin_unlock(&fs_info->ordered_root_lock); |
199c2a9c MX |
4089 | btrfs_destroy_ordered_extents(root); |
4090 | ||
2a85d9ca LB |
4091 | cond_resched(); |
4092 | spin_lock(&fs_info->ordered_root_lock); | |
199c2a9c MX |
4093 | } |
4094 | spin_unlock(&fs_info->ordered_root_lock); | |
acce952b | 4095 | } |
4096 | ||
35a3621b | 4097 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
2ff7e61e | 4098 | struct btrfs_fs_info *fs_info) |
acce952b | 4099 | { |
4100 | struct rb_node *node; | |
4101 | struct btrfs_delayed_ref_root *delayed_refs; | |
4102 | struct btrfs_delayed_ref_node *ref; | |
4103 | int ret = 0; | |
4104 | ||
4105 | delayed_refs = &trans->delayed_refs; | |
4106 | ||
4107 | spin_lock(&delayed_refs->lock); | |
d7df2c79 | 4108 | if (atomic_read(&delayed_refs->num_entries) == 0) { |
cfece4db | 4109 | spin_unlock(&delayed_refs->lock); |
0b246afa | 4110 | btrfs_info(fs_info, "delayed_refs has NO entry"); |
acce952b | 4111 | return ret; |
4112 | } | |
4113 | ||
d7df2c79 JB |
4114 | while ((node = rb_first(&delayed_refs->href_root)) != NULL) { |
4115 | struct btrfs_delayed_ref_head *head; | |
0e0adbcf | 4116 | struct rb_node *n; |
e78417d1 | 4117 | bool pin_bytes = false; |
acce952b | 4118 | |
d7df2c79 JB |
4119 | head = rb_entry(node, struct btrfs_delayed_ref_head, |
4120 | href_node); | |
4121 | if (!mutex_trylock(&head->mutex)) { | |
d278850e | 4122 | refcount_inc(&head->refs); |
d7df2c79 | 4123 | spin_unlock(&delayed_refs->lock); |
eb12db69 | 4124 | |
d7df2c79 | 4125 | mutex_lock(&head->mutex); |
e78417d1 | 4126 | mutex_unlock(&head->mutex); |
d278850e | 4127 | btrfs_put_delayed_ref_head(head); |
d7df2c79 JB |
4128 | spin_lock(&delayed_refs->lock); |
4129 | continue; | |
4130 | } | |
4131 | spin_lock(&head->lock); | |
0e0adbcf JB |
4132 | while ((n = rb_first(&head->ref_tree)) != NULL) { |
4133 | ref = rb_entry(n, struct btrfs_delayed_ref_node, | |
4134 | ref_node); | |
d7df2c79 | 4135 | ref->in_tree = 0; |
0e0adbcf JB |
4136 | rb_erase(&ref->ref_node, &head->ref_tree); |
4137 | RB_CLEAR_NODE(&ref->ref_node); | |
1d57ee94 WX |
4138 | if (!list_empty(&ref->add_list)) |
4139 | list_del(&ref->add_list); | |
d7df2c79 JB |
4140 | atomic_dec(&delayed_refs->num_entries); |
4141 | btrfs_put_delayed_ref(ref); | |
e78417d1 | 4142 | } |
d7df2c79 JB |
4143 | if (head->must_insert_reserved) |
4144 | pin_bytes = true; | |
4145 | btrfs_free_delayed_extent_op(head->extent_op); | |
4146 | delayed_refs->num_heads--; | |
4147 | if (head->processing == 0) | |
4148 | delayed_refs->num_heads_ready--; | |
4149 | atomic_dec(&delayed_refs->num_entries); | |
d7df2c79 | 4150 | rb_erase(&head->href_node, &delayed_refs->href_root); |
d278850e | 4151 | RB_CLEAR_NODE(&head->href_node); |
d7df2c79 JB |
4152 | spin_unlock(&head->lock); |
4153 | spin_unlock(&delayed_refs->lock); | |
4154 | mutex_unlock(&head->mutex); | |
acce952b | 4155 | |
d7df2c79 | 4156 | if (pin_bytes) |
d278850e JB |
4157 | btrfs_pin_extent(fs_info, head->bytenr, |
4158 | head->num_bytes, 1); | |
4159 | btrfs_put_delayed_ref_head(head); | |
acce952b | 4160 | cond_resched(); |
4161 | spin_lock(&delayed_refs->lock); | |
4162 | } | |
4163 | ||
4164 | spin_unlock(&delayed_refs->lock); | |
4165 | ||
4166 | return ret; | |
4167 | } | |
4168 | ||
143bede5 | 4169 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) |
acce952b | 4170 | { |
4171 | struct btrfs_inode *btrfs_inode; | |
4172 | struct list_head splice; | |
4173 | ||
4174 | INIT_LIST_HEAD(&splice); | |
4175 | ||
eb73c1b7 MX |
4176 | spin_lock(&root->delalloc_lock); |
4177 | list_splice_init(&root->delalloc_inodes, &splice); | |
acce952b | 4178 | |
4179 | while (!list_empty(&splice)) { | |
eb73c1b7 MX |
4180 | btrfs_inode = list_first_entry(&splice, struct btrfs_inode, |
4181 | delalloc_inodes); | |
acce952b | 4182 | |
4183 | list_del_init(&btrfs_inode->delalloc_inodes); | |
df0af1a5 MX |
4184 | clear_bit(BTRFS_INODE_IN_DELALLOC_LIST, |
4185 | &btrfs_inode->runtime_flags); | |
eb73c1b7 | 4186 | spin_unlock(&root->delalloc_lock); |
acce952b | 4187 | |
4188 | btrfs_invalidate_inodes(btrfs_inode->root); | |
b216cbfb | 4189 | |
eb73c1b7 | 4190 | spin_lock(&root->delalloc_lock); |
acce952b | 4191 | } |
4192 | ||
eb73c1b7 MX |
4193 | spin_unlock(&root->delalloc_lock); |
4194 | } | |
4195 | ||
4196 | static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) | |
4197 | { | |
4198 | struct btrfs_root *root; | |
4199 | struct list_head splice; | |
4200 | ||
4201 | INIT_LIST_HEAD(&splice); | |
4202 | ||
4203 | spin_lock(&fs_info->delalloc_root_lock); | |
4204 | list_splice_init(&fs_info->delalloc_roots, &splice); | |
4205 | while (!list_empty(&splice)) { | |
4206 | root = list_first_entry(&splice, struct btrfs_root, | |
4207 | delalloc_root); | |
4208 | list_del_init(&root->delalloc_root); | |
4209 | root = btrfs_grab_fs_root(root); | |
4210 | BUG_ON(!root); | |
4211 | spin_unlock(&fs_info->delalloc_root_lock); | |
4212 | ||
4213 | btrfs_destroy_delalloc_inodes(root); | |
4214 | btrfs_put_fs_root(root); | |
4215 | ||
4216 | spin_lock(&fs_info->delalloc_root_lock); | |
4217 | } | |
4218 | spin_unlock(&fs_info->delalloc_root_lock); | |
acce952b | 4219 | } |
4220 | ||
2ff7e61e | 4221 | static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, |
acce952b | 4222 | struct extent_io_tree *dirty_pages, |
4223 | int mark) | |
4224 | { | |
4225 | int ret; | |
acce952b | 4226 | struct extent_buffer *eb; |
4227 | u64 start = 0; | |
4228 | u64 end; | |
acce952b | 4229 | |
4230 | while (1) { | |
4231 | ret = find_first_extent_bit(dirty_pages, start, &start, &end, | |
e6138876 | 4232 | mark, NULL); |
acce952b | 4233 | if (ret) |
4234 | break; | |
4235 | ||
91166212 | 4236 | clear_extent_bits(dirty_pages, start, end, mark); |
acce952b | 4237 | while (start <= end) { |
0b246afa JM |
4238 | eb = find_extent_buffer(fs_info, start); |
4239 | start += fs_info->nodesize; | |
fd8b2b61 | 4240 | if (!eb) |
acce952b | 4241 | continue; |
fd8b2b61 | 4242 | wait_on_extent_buffer_writeback(eb); |
acce952b | 4243 | |
fd8b2b61 JB |
4244 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, |
4245 | &eb->bflags)) | |
4246 | clear_extent_buffer_dirty(eb); | |
4247 | free_extent_buffer_stale(eb); | |
acce952b | 4248 | } |
4249 | } | |
4250 | ||
4251 | return ret; | |
4252 | } | |
4253 | ||
2ff7e61e | 4254 | static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, |
acce952b | 4255 | struct extent_io_tree *pinned_extents) |
4256 | { | |
4257 | struct extent_io_tree *unpin; | |
4258 | u64 start; | |
4259 | u64 end; | |
4260 | int ret; | |
ed0eaa14 | 4261 | bool loop = true; |
acce952b | 4262 | |
4263 | unpin = pinned_extents; | |
ed0eaa14 | 4264 | again: |
acce952b | 4265 | while (1) { |
4266 | ret = find_first_extent_bit(unpin, 0, &start, &end, | |
e6138876 | 4267 | EXTENT_DIRTY, NULL); |
acce952b | 4268 | if (ret) |
4269 | break; | |
4270 | ||
af6f8f60 | 4271 | clear_extent_dirty(unpin, start, end); |
2ff7e61e | 4272 | btrfs_error_unpin_extent_range(fs_info, start, end); |
acce952b | 4273 | cond_resched(); |
4274 | } | |
4275 | ||
ed0eaa14 | 4276 | if (loop) { |
0b246afa JM |
4277 | if (unpin == &fs_info->freed_extents[0]) |
4278 | unpin = &fs_info->freed_extents[1]; | |
ed0eaa14 | 4279 | else |
0b246afa | 4280 | unpin = &fs_info->freed_extents[0]; |
ed0eaa14 LB |
4281 | loop = false; |
4282 | goto again; | |
4283 | } | |
4284 | ||
acce952b | 4285 | return 0; |
4286 | } | |
4287 | ||
c79a1751 LB |
4288 | static void btrfs_cleanup_bg_io(struct btrfs_block_group_cache *cache) |
4289 | { | |
4290 | struct inode *inode; | |
4291 | ||
4292 | inode = cache->io_ctl.inode; | |
4293 | if (inode) { | |
4294 | invalidate_inode_pages2(inode->i_mapping); | |
4295 | BTRFS_I(inode)->generation = 0; | |
4296 | cache->io_ctl.inode = NULL; | |
4297 | iput(inode); | |
4298 | } | |
4299 | btrfs_put_block_group(cache); | |
4300 | } | |
4301 | ||
4302 | void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, | |
2ff7e61e | 4303 | struct btrfs_fs_info *fs_info) |
c79a1751 LB |
4304 | { |
4305 | struct btrfs_block_group_cache *cache; | |
4306 | ||
4307 | spin_lock(&cur_trans->dirty_bgs_lock); | |
4308 | while (!list_empty(&cur_trans->dirty_bgs)) { | |
4309 | cache = list_first_entry(&cur_trans->dirty_bgs, | |
4310 | struct btrfs_block_group_cache, | |
4311 | dirty_list); | |
4312 | if (!cache) { | |
0b246afa | 4313 | btrfs_err(fs_info, "orphan block group dirty_bgs list"); |
c79a1751 LB |
4314 | spin_unlock(&cur_trans->dirty_bgs_lock); |
4315 | return; | |
4316 | } | |
4317 | ||
4318 | if (!list_empty(&cache->io_list)) { | |
4319 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4320 | list_del_init(&cache->io_list); | |
4321 | btrfs_cleanup_bg_io(cache); | |
4322 | spin_lock(&cur_trans->dirty_bgs_lock); | |
4323 | } | |
4324 | ||
4325 | list_del_init(&cache->dirty_list); | |
4326 | spin_lock(&cache->lock); | |
4327 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
4328 | spin_unlock(&cache->lock); | |
4329 | ||
4330 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4331 | btrfs_put_block_group(cache); | |
4332 | spin_lock(&cur_trans->dirty_bgs_lock); | |
4333 | } | |
4334 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4335 | ||
4336 | while (!list_empty(&cur_trans->io_bgs)) { | |
4337 | cache = list_first_entry(&cur_trans->io_bgs, | |
4338 | struct btrfs_block_group_cache, | |
4339 | io_list); | |
4340 | if (!cache) { | |
0b246afa | 4341 | btrfs_err(fs_info, "orphan block group on io_bgs list"); |
c79a1751 LB |
4342 | return; |
4343 | } | |
4344 | ||
4345 | list_del_init(&cache->io_list); | |
4346 | spin_lock(&cache->lock); | |
4347 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
4348 | spin_unlock(&cache->lock); | |
4349 | btrfs_cleanup_bg_io(cache); | |
4350 | } | |
4351 | } | |
4352 | ||
49b25e05 | 4353 | void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, |
2ff7e61e | 4354 | struct btrfs_fs_info *fs_info) |
49b25e05 | 4355 | { |
2ff7e61e | 4356 | btrfs_cleanup_dirty_bgs(cur_trans, fs_info); |
c79a1751 LB |
4357 | ASSERT(list_empty(&cur_trans->dirty_bgs)); |
4358 | ASSERT(list_empty(&cur_trans->io_bgs)); | |
4359 | ||
2ff7e61e | 4360 | btrfs_destroy_delayed_refs(cur_trans, fs_info); |
49b25e05 | 4361 | |
4a9d8bde | 4362 | cur_trans->state = TRANS_STATE_COMMIT_START; |
0b246afa | 4363 | wake_up(&fs_info->transaction_blocked_wait); |
49b25e05 | 4364 | |
4a9d8bde | 4365 | cur_trans->state = TRANS_STATE_UNBLOCKED; |
0b246afa | 4366 | wake_up(&fs_info->transaction_wait); |
49b25e05 | 4367 | |
ccdf9b30 JM |
4368 | btrfs_destroy_delayed_inodes(fs_info); |
4369 | btrfs_assert_delayed_root_empty(fs_info); | |
49b25e05 | 4370 | |
2ff7e61e | 4371 | btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, |
49b25e05 | 4372 | EXTENT_DIRTY); |
2ff7e61e | 4373 | btrfs_destroy_pinned_extent(fs_info, |
0b246afa | 4374 | fs_info->pinned_extents); |
49b25e05 | 4375 | |
4a9d8bde MX |
4376 | cur_trans->state =TRANS_STATE_COMPLETED; |
4377 | wake_up(&cur_trans->commit_wait); | |
49b25e05 JM |
4378 | } |
4379 | ||
2ff7e61e | 4380 | static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) |
acce952b | 4381 | { |
4382 | struct btrfs_transaction *t; | |
acce952b | 4383 | |
0b246afa | 4384 | mutex_lock(&fs_info->transaction_kthread_mutex); |
acce952b | 4385 | |
0b246afa JM |
4386 | spin_lock(&fs_info->trans_lock); |
4387 | while (!list_empty(&fs_info->trans_list)) { | |
4388 | t = list_first_entry(&fs_info->trans_list, | |
724e2315 JB |
4389 | struct btrfs_transaction, list); |
4390 | if (t->state >= TRANS_STATE_COMMIT_START) { | |
9b64f57d | 4391 | refcount_inc(&t->use_count); |
0b246afa | 4392 | spin_unlock(&fs_info->trans_lock); |
2ff7e61e | 4393 | btrfs_wait_for_commit(fs_info, t->transid); |
724e2315 | 4394 | btrfs_put_transaction(t); |
0b246afa | 4395 | spin_lock(&fs_info->trans_lock); |
724e2315 JB |
4396 | continue; |
4397 | } | |
0b246afa | 4398 | if (t == fs_info->running_transaction) { |
724e2315 | 4399 | t->state = TRANS_STATE_COMMIT_DOING; |
0b246afa | 4400 | spin_unlock(&fs_info->trans_lock); |
724e2315 JB |
4401 | /* |
4402 | * We wait for 0 num_writers since we don't hold a trans | |
4403 | * handle open currently for this transaction. | |
4404 | */ | |
4405 | wait_event(t->writer_wait, | |
4406 | atomic_read(&t->num_writers) == 0); | |
4407 | } else { | |
0b246afa | 4408 | spin_unlock(&fs_info->trans_lock); |
724e2315 | 4409 | } |
2ff7e61e | 4410 | btrfs_cleanup_one_transaction(t, fs_info); |
4a9d8bde | 4411 | |
0b246afa JM |
4412 | spin_lock(&fs_info->trans_lock); |
4413 | if (t == fs_info->running_transaction) | |
4414 | fs_info->running_transaction = NULL; | |
acce952b | 4415 | list_del_init(&t->list); |
0b246afa | 4416 | spin_unlock(&fs_info->trans_lock); |
acce952b | 4417 | |
724e2315 | 4418 | btrfs_put_transaction(t); |
2ff7e61e | 4419 | trace_btrfs_transaction_commit(fs_info->tree_root); |
0b246afa | 4420 | spin_lock(&fs_info->trans_lock); |
724e2315 | 4421 | } |
0b246afa JM |
4422 | spin_unlock(&fs_info->trans_lock); |
4423 | btrfs_destroy_all_ordered_extents(fs_info); | |
ccdf9b30 JM |
4424 | btrfs_destroy_delayed_inodes(fs_info); |
4425 | btrfs_assert_delayed_root_empty(fs_info); | |
2ff7e61e | 4426 | btrfs_destroy_pinned_extent(fs_info, fs_info->pinned_extents); |
0b246afa JM |
4427 | btrfs_destroy_all_delalloc_inodes(fs_info); |
4428 | mutex_unlock(&fs_info->transaction_kthread_mutex); | |
acce952b | 4429 | |
4430 | return 0; | |
4431 | } | |
4432 | ||
c6100a4b JB |
4433 | static struct btrfs_fs_info *btree_fs_info(void *private_data) |
4434 | { | |
4435 | struct inode *inode = private_data; | |
4436 | return btrfs_sb(inode->i_sb); | |
4437 | } | |
4438 | ||
e8c9f186 | 4439 | static const struct extent_io_ops btree_extent_io_ops = { |
4d53dddb | 4440 | /* mandatory callbacks */ |
0b86a832 | 4441 | .submit_bio_hook = btree_submit_bio_hook, |
4d53dddb | 4442 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
239b14b3 CM |
4443 | /* note we're sharing with inode.c for the merge bio hook */ |
4444 | .merge_bio_hook = btrfs_merge_bio_hook, | |
20a7db8a | 4445 | .readpage_io_failed_hook = btree_io_failed_hook, |
c6100a4b JB |
4446 | .set_range_writeback = btrfs_set_range_writeback, |
4447 | .tree_fs_info = btree_fs_info, | |
4d53dddb DS |
4448 | |
4449 | /* optional callbacks */ | |
0da5468f | 4450 | }; |